Professional Documents
Culture Documents
Dissertation submitted to the Faculty of the Virginia Polytechnic Institute and State University in
partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in
Electrical Engineering
_______________________________
Fred C. Lee, Chairman
_______________________________ _______________________________
Dushan Boroyevich Dan Y. Chen
_______________________________ _______________________________
Douglas Lindner Werner Kohler
Front-end converters with power factor correction (PFC) capability are widely used in distributed
power systems (DPSs). Most of the front-end converters are implemented using a two-stage
approach, which consists of a PFC stage followed by a DC/DC converter. The purpose of the
front-end converter is to regulate the DC output voltage, supply all the load converters connected
to the distributed bus, guarantee current sharing, and charge a bank of batteries to provide
backup energy when the power grid breaks down.
One of the main concerns of the power supply industry is to obtain a front-end converter with a
low-cost PFC stage, while still complying with required harmonic standards, especially for high-
power three-phase applications. Having this statement in mind, the main objective of this
dissertation is to study front-end converters for DPS applications with PFC to meet harmonic
standards, while still maintaining low cost and performance indices.
To realize the many aforementioned objectives, this dissertation is divided into two main parts:
(1) two-stage front-end converters suitable for telecom applications, and (2) single-stage low-
cost AC/DC converters suitable for mainframe computers and server applications. The use of
discontinuous conduction mode (DCM) boost rectifiers is extensively explored to achieve
simplicity, while reducing the cost for DPS applications. Interleaving of DCM boost rectifiers is
also explored as an alternative approach to further reduce the system cost by reducing the
filtering requirements. All the solutions discussed are implemented for 3kW applications, while
6kW is obtained by interleaving two converters
Acknowledgments1
I would like to express the most sincere appreciation to those who made this work possible:
professors, friends and family.
I would like to express my great admiration of Dr. Lee, a living legend in the power electronics
community. I thank him for his support, belief, patience, fairness, and for his feedback. He
taught me something beyond just techniques for solving problems. He taught me attitude,
initiative, and how to have passion for what we believe. I have to thank him for the many
opportunities he has given me over the years. I came to Virginia Tech as a PhD student and now
I work as the technical coordinator for the Center for Power Electronics Systems (CPES).
Although many did not believe that I could deal with this pressure, he was the one who always
gave me all the support that I always needed to do my job. Thank you Dr. Lee for your support
and friendship. I think I am ready for another site visit!
To Dr. B, the power point guy, who is another person very close to what I do on a daily basis
(which are power point presentations). We have had a lot of fun together, from the IPEMS
quarterly review meetings to the site visits. Oh yeah, I couldn’t forget the annual retreat in
Florida, which I hope this year will be in Key West. Thank you for your support and for being so
cheerful on every occasion. It does not matter the situation: if something has been screwed up
(not by me, of course), he will always find a way out.
Thanks also to Dr. Chen, my former professor and a member of my advisory committee. He is
the EMI expert in our group and has a very good sense of humor. And to Dr. Lindner, who was
also my professor and is a good wine drinker. He taught me multivariable control, but I have to
tell you that I am still trying to figure out what he was talking about in those classes. Thanks also
to Dr. Kohler from the math department, who has served as a member of my committee.
1
This work was supported by the Brazilian Council for Scientific and Technologic Development (CNPq), the
Federal University of Paraná (UFPR), and by the ERC Program of the National Science Foundation (NSF) under
Award Number EEC-9731677.
iii
I would also like to express my appreciation to Dr. van Wyk, an exceptional mind driving the
technology efforts of CPES, and a very insightful storyteller. I would like to acknowledge Prof.
Bob Lorenz from the University of Wisconsin at Madison for his support since the time when I
first assumed the CPES technical coordination.
There are many bright-minded individuals who are part of CPES and who are having a huge
impact on my formation: Profs. Lipo and Jahns from the University of Wisconsin, and Profs.
Chow and Gutmann from the Rensselaer Polytechnic Institute.
I also have to express my deepest admiration to Prof. Ivo Barbi, my former advisor in my master
degree and the person responsible for introducing the first power electronics program into the
Brazilian Academia. He is a friend whose steps I will try to follow during the course of my life.
My gratitude also goes to CPES staff, who have been so supportive of the type of work that I do
every day. To Teresa Shaw - “nossa mainha,” which literally means “little mother” in Portuguese
- a great friend who we met in Blacksburg. To Beth Tranter who is so professional about
everything she does. To Ann Craig, the toughest person in CPES – somebody has to keep
everyone on track! To Linda Gallagher, Trish Rose, Lesli Farmer, Linda Long, Marianne
Hawthorne and Amy Shea for editing the final edition of this dissertation. I also want to thank
Dan Huff, Bob Martin, Steve Chen, Jamie Evans, Gary Kerr, Dr. Liang and Mike King for their
support.
To Francisco Canales, a friend who I met in the first day when I came to CPES. He has not only
been a great friend, but also a great project mate. He is also coming out of the pipeline very soon.
I wish him all the best, and know that our friendship will last forever. The only problem is that
Mexico lost to the U.S. in the World Cup!
I have met so many good people in CPES. I have to apologize if I forget to cite some of you.
Thanks to Siriroj Sirisukprasert, Sudip Mazumder, Eric Hertz, Zhao Qun, Sriram
Chandrasekaram, Yong Li, Henry Zhang, Xu Peng, Sam Ye, Ray-Lee Lin, Yuxin Li, Kalyan
Siddabattula (when are you going to return my CDs?), Naveen Yadlapalli, Feng Feng Tao, Shatil
Haque, Nikola Celanovic, Simon Wen, JinJun Liu, Dimos Katsis, Johan Strydom, Wei Dong, Y.
Pang, Evan Sewall, Xu Ming, Deng-Ming Peng, Bing Lu, Bo Yang, Roger Chen, Gary Yao, Wei
Jia, Jonah Chen, Jeremy Ferrel, Jerry Francis, Jinghong Guo, Han Chong, Xudong Huang,
Changrong Liu, Troy Negaard, Yang Qiu, Carl Tinsley, Alex Uan-Zo-Li (we still need to finish
iv
that motor control class), Shen Wang, Wang Shuo, Mountain, Aaron Xu, Lyiu Yang, Huijie Yu,
Zach Zhang, Lingying Zhao, Jinghai Zhou and Yuancheng Ren.
Thanks to the support from Jessy Alves Pinheiro, José Airton de Souza, Alcina Taitson Queiroz
and Nelson Prugner. They are the staff of the Brazilian Council for Scientific and Technologic
Development (CNPq). I also want to thank Vera Bichara, a great friend who we left in Brazil.
You have been so important during these years, and Leandra also has a great admiration for you.
Special thanks to our Catalan friends Sergio Monge-Busquests (I love that Sangria), Josep Pou,
Montse Olive (we miss your cooking), and Josep Bordonau (it’s about time again for us to
prepare that paella).
A huge thanks to my fellow Brazilian friends Luciano da Silva, Marcelo Cavalcanti, Paulo
Cardieri (already in Brazil), Rodrigo Marques, Carmen and Valder Steffen Jr (one of the most
prominent Brazilian researchers).
To Leonardo Serpa who spent some time in CPES and helped me in my dissertation. Without
your help, I would not have been able to finish this work. Thank you for the simulations and
prototyping. I wish you a very successful career, and welcome you to the power electronics
community.
Thanks to Arnaldo José Perin, a former Brazilian professor who motivated us to pursue my
degree abroad. He and his wife Carla know some good things about life, food and wine. To our
family and friends in Brazil: Marcelo, Marly, Marcos, Márcia, Leandro, Alexandre, Maria Luiza
(she is coming!), Dona Luci, Tia Amélia, Tia Alice, Lourenço and Rosana, Guilherme, Adriano,
Trovão and Juliana, Naná, Paulo Gaidzinski, Renes, Darizon and Juliana, Kleber, Carla, Tia
Ruth, Vó Lina, Vó Leonor, Vô de Cianorte, Sérgio e Mirtes, and so many others who helped me
along the way. Thank you very much.
I would like to specially thank my father Rubens Inácio Barbosa and my mother Dirce Lourdes
Mantovanelli Barbosa for their support since the beginning. I will never forget all their incentive,
despite all the financial odds that we lived through together during the 80s. I would also like to
dedicate this dissertation to my grandfather José Osório Barbosa, who passed away long ago, but
taught me honesty and modesty, which are the essential ingredients to a simple life.
v
I dedicate this achievement to Leandra Machado
It would not have been possible without your support, encouragement and love. Thank you for
being with me in those dark moments, including those when I had decided to drop out and return
to Brazil on the first available flight.
Thank you for making my life a bit spicier. Thank you for being you, so cheerful and emotional.
Without you, I would be a lost soul.
vi
Table of Contents
vii
2.8. Two-Stage Front-End Converter Using the Interleaved Single-Switch DCM Boost
Rectifier As the Front-End PFC ................................................................................. 67
2.8.1. DC/DC Topology for 800V Bus Voltage Applications........................................ 68
2.8.2. Interfacing PFC and DC/DC Converters .............................................................. 73
2.9. Benchmarking ............................................................................................................... 76
2.10. Conclusion .................................................................................................................... 78
viii
4.3.3. Design Guidelines and Example......................................................................... 132
4.3.4. Harmonic Distortion of the Input Current .......................................................... 135
4.3.5. Experimental Results .......................................................................................... 136
4.4. Single-Stage Three-Level Asymmetrical (TL-AS) Front-End Converter .................. 141
4.4.1. Circuit Description and Operation ...................................................................... 142
4.4.2. Analysis of the TL-PS AC/DC Converter .......................................................... 145
4.4.3. Design Guidelines and Example......................................................................... 151
4.4.4. Experimental Results and Comparisons ............................................................. 155
4.5. Interleaved Single-Stage AC/DC Converters ............................................................. 159
4.6. Output Current Ripple Cancellation in Single-Stage Converters ............................... 164
4.7. C-Message and Psophometric Noise Levels in Single-Stage Front-End Converters . 165
4.7.1. Voltage Loop Bandwidth of Single-Stage Front-End Converters ...................... 167
4.7.2. Installation and Battery String Impedances ........................................................ 171
4.8. Benchmarking ............................................................................................................. 173
4.9. Conclusion .................................................................................................................. 176
REFERENCES.......................................................................................................................... 222
ix
List of Figures
x
Fig. 2-11. Two-channel interleaved DCM boost rectifiers: (a) configuration, (b) results without
harmonic injection at Vin=220V (line-to-neutral), Po=8kW, fs=40kHz, and Vo=800V
(THD=12.7%), and (c) results with harmonic injection (THD=10.8%). All current traces are
20A/div and all voltage traces are 200V/div................................................................................. 50
Fig. 2-12. (a) Load sharing and (b) efficiency. ............................................................................. 51
Fig. 2-13. (a) Low-order harmonic comparison, (b) high-frequency spectrum for each control
strategy and (c) equivalent high-frequency spectrum when a 9kHz window is considered in the
spectrum calculation (quadratic sum). .......................................................................................... 54
Fig. 2-14. The impact of phase-shift error on the amplitude of the harmonic at the switching
frequency for two interleaved rectifiers operated with fixed frequency and fixed duty cycle
control. .......................................................................................................................................... 55
Fig. 2-15. (a) EMI standards for conducted noise (average limits) and (b) equivalent DM filter
for one phase. ................................................................................................................................ 56
Fig. 2-16. Single-phase conducted EMI simulation setup. ........................................................... 57
Fig. 2-17. Conducted EMI noise simulation setup for three-phase equipment. ........................... 58
Fig. 2-18. Input current ripple harmonics: (a) single-phase CCM boost rectifier (2kW) and (b)
VIENNA rectifier (6kW total power – 2kW per phase). .............................................................. 59
Fig. 2-19. Input current ripple harmonics: (a) one non-interleaved DCM boost rectifier (6kW),
and (b) two-interleaved rectifiers (6kW). ..................................................................................... 60
Fig. 2-20. Boost inductor design results: (a) boost inductance, (b) core weight, (c) Cu weight, and
(d) combined core + Cu weight per phase. ................................................................................... 63
Fig. 2-21. Filter size for VDE 0871 Class B: (a) filter inductance L1, (b) filter inductance L3, (c)
filter core weight to implement Ld+L1+L3 and (d) Core + Cu weight per phase required to
implement Ld+L1+L3..................................................................................................................... 65
Fig. 2-22. Filter size for CISPR 22 Class B: (a) filter inductance L1, (b) filter inductance L3, (c)
filter core weight to implement Ld+L1+L3, and (d) Core + Cu weight per phase required to
implement Ld+L1+L3..................................................................................................................... 67
Fig. 2-23. (a) Combined filter and boost inductor core weight per phase and (b) total core + Cu
weight needed to implement the boost and filter inductors per phase.......................................... 68
Fig. 2-24. DC/DC power conversion topologies: (a) full-bridge, (b) dual-bridge and (c) three-
level............................................................................................................................................... 69
Fig. 2-25. Theoretical efficiency calculated for full-bridge and three-level converters. .............. 71
Fig. 2-26. (a) Three-level ZVZCS DC/DC converter used in the implementation of the two-stage
front-end converter, (b) experimental results for vab, vs and iLlk (20A/div) at Po=5kW, Vbus=800V,
Vout=52V and fs=100kHz, and (c) ZVS transition for the outer switches..................................... 73
Fig. 2-27. Front-end converters using interleaved DCM boost rectifiers: (a) common
intermediate bus voltage, and (b) using two DC/DC converters. ................................................. 75
Fig. 2-28. System efficiency at 6kW, including EMI filter. ......................................................... 76
Fig. 3-1. Two-switch three-level PFC circuit. .............................................................................. 81
xi
Fig. 3-2. Operating stages: (a) first stage (to, t1), (b) second stage (t1, t2), (c) third stage (t2, t3),
and (d) fourth stage (t3, t4)............................................................................................................. 83
Fig. 3-3. Boost inductor current waveforms. ................................................................................ 84
Fig. 3-4. Control of voltages across C1 and C2: (a) control scheme and (b) simulation results.... 85
Fig. 3-5. Possible applications: (a) using two split DC/DC converters and (b) using a three-level
DC/DC topology. .......................................................................................................................... 86
Fig. 3-6. Using the DC/DC converters to control the voltage imbalance across the DC capacitors
of the three-level DCM boost rectifier: (a) circuit diagram and (b) voltages across C1 and C2. .. 88
Fig. 3-7. Voltage gain versus normalized output current at heavy load. ...................................... 90
Fig. 3-8. Experimental results: (a) boost inductor current (10A/div) at 3kW and Vin=180V and
(b) voltage across one of the switches. ......................................................................................... 91
Fig. 3-9. (a) Efficiency comparison and (b) THD. ....................................................................... 93
Fig. 3-10. The two-switch three-level DCM boost rectifier using the harmonic injection method:
(a) implementation and (b) THD. ................................................................................................. 95
Fig. 3-11. (a) DCM interleaved two-switch boost rectifiers, (b) high-frequency harmonics
without interleaving, and (c) high-frequency harmonics with interleaved operation................... 97
Fig. 3-12. Interleaved current at 6.3kW (iL11+iL12) (10A/div all traces) at three different input
phase voltages. .............................................................................................................................. 98
Fig. 3-13. Filter size for VDE 0871 Class B: (a) filter inductance L1, (b) filter inductance L3, (c)
filter core weight to implement Ld+L1+L3, and (d) Core + Cu weight per phase required to
implement Ld+L1+L3................................................................................................................... 100
Fig. 3-14. Filter size for CISPR 22 Class B: (a) filter inductance L1, (b) filter inductance L3, (c)
filter core weight to implement Ld+L1+L3, and (d) Core + Cu weight per phase required to
implement Ld+L1+L3................................................................................................................... 101
Fig. 3-15. (a) Combined filter and boost inductor core weight per phase and (b) total core + Cu
weight needed to implement the boost and filter inductors per phase........................................ 102
Fig. 3-16. Equivalent model for each boost inductor: (a) single-phase CCM boost and VIENNA
rectifiers and (b) two-switch three-level interleaved system. ..................................................... 104
Fig. 3-17. Device assembly......................................................................................................... 105
Fig. 3-18. Parasitic components: (a) single-phase CCM boost rectifier, (b) VIENNA rectifier and
(c) two-switch three-level interleaved system. ........................................................................... 107
Fig. 3-19. Equivalent high-frequency noise path: (a) single-phase and (b) three-phase. ........... 109
Fig. 3-20. CM noise generated by the various converters: (a) CCM boost, (b) VIENNA and (c)
interleaved................................................................................................................................... 111
Fig. 3-21. Device stress: (a) average switch current, (b) RMS switch current and (c) RMS bus
capacitor current.......................................................................................................................... 112
Fig. 3-22. Efficiency comparison................................................................................................ 113
xii
Fig. 3-23. (a) Two-stage front-end converter and (b) experimental efficiency comparison....... 115
Fig. 4-1. Two-stage front-end converter using three-level topologies........................................ 121
Fig. 4-2. Synthesis of TL-PS single-stage AC/DC converter: (a) eliminating the PFC cell, and (b)
reconnecting the artificial neutral point. ..................................................................................... 123
Fig. 4-3. Main waveforms........................................................................................................... 124
Fig. 4-4. Equivalent circuits assumed during the operating stages............................................. 127
Fig. 4-5. Illustration of the boost inductor current waveform for half of a line period. ............. 128
Fig. 4-6. Normalized bus voltage gain........................................................................................ 130
Fig. 4-7. Normalized DC output voltage as a function of normalized DC output current.......... 131
Fig. 4-8. Boost inductance as a function of the switching frequency for DCM operation. ........ 134
Fig. 4-9. Calculated maximum intermediate bus voltage stress as a function of inductance ratio.
..................................................................................................................................................... 135
Fig. 4-10. THD: (a) with third harmonic, and (b) without third harmonic. ................................ 137
Fig. 4-11. Experimental waveforms: (a) input voltage (100V/div) and filtered input current
(5A/div), (b) voltage vab (250V/div) and primary current (10A/div), and (c) ZVS waveforms
(voltages: 200V/div and currents: 20A/div). .............................................................................. 138
Fig. 4-12. Experimental results: (a) intermediate bus voltage, (b) THD, and (c) converter
efficiency measured for both designs. ........................................................................................ 140
Fig. 4-13. Main current path through the power switches. ......................................................... 141
Fig. 4-14. Three-level asymmetrical DC/DC converter: (a) topology and (b) waveforms......... 142
Fig. 4-15. Synthesis process: (a) two-stage approach, and (b) TL-AS single-stage AC/DC
converter. .................................................................................................................................... 143
Fig. 4-16. TL-AS converter main waveforms............................................................................. 144
Fig. 4-17. Operating stages. ........................................................................................................ 146
Fig. 4-18. Normalized bus voltage gain for the TL-AS AC/DC converter................................. 148
Fig. 4-19. DC-side DCM operation: (a) inductor current through Lo, (b) magnetizing stage for Lo,
(c) resetting stage for Lo and reversing current polarity in Lr, and (d) final resetting stage for Lo.
..................................................................................................................................................... 152
Fig. 4-20. Boost inductance as a function of the switching frequency for DCM operation. ...... 155
Fig. 4-21. Experimental results: (a) drain voltage and gate signal for S1 at 3kW and Vin=180V,
(b) drain voltage and gate signal for S2 at 3kW and Vin=180V, (c) primary current at 3kW for
three input voltages, and (d) filtered input current at 3kW for three input voltages................... 158
Fig. 4-22. Results and comparisons: (a) intermediate bus voltage stress,(b) DC blocking
capacitor, (c) efficiency comparison, and (d) THD comparison. ............................................... 159
Fig. 4-23. Interleaved single-stage converters: (a) TL-PS and (b) TL-AS. ................................ 160
xiii
Fig. 4-24. Harmonics of switching frequency: (a) non-interleaved 6kW TL-PS, (b) interleaved
6kW TL-PS, (c) non-interleaved 6kW TL-AS, and (d) interleaved 6kW TL-AS. ..................... 161
Fig. 4-25. Design results: (a) boost inductance, combined boost inductor core + Cu weight per
phase, (c) filter inductance L1, (d) filter inductance L3, (e) filter core weight needed to implement
Ld+L1+L3, and (d) Core + Cu weight per phase required to implement Ld+L1+L3. ................... 162
Fig. 4-26. Combined filter and boost inductor core weight per phase and (b) total core + Cu
weight needed to implement the boost and filter inductors per phase........................................ 164
Fig. 4-27. Output current ripple cancellation in interleaved TL-PS converters: (a) voltage vab,
rectified secondary voltage, and (c) output inductor currents..................................................... 166
Fig. 4-28. Output current ripple cancellation in interleaved TL-AS converters: (a) voltage vab,
rectified secondary voltage, and (c) interleaved output inductor currents.................................. 166
Fig. 4-29. C-message weighing factor. ....................................................................................... 168
Fig. 4-30. Small-signal model of the equivalent three-level ZVS phase-shift DC/DC converter
representing the DC section of the single-stage TL-PS converter.............................................. 169
Fig. 4-31. Simulated input current THD as a function of the crossover frequency of the single-
stage TL-PS converter................................................................................................................. 171
Fig. 4-32. (a) Installation and battery string configuration and (b) equivalent circuit................ 172
Fig. 4-33. (a) C-message output voltage noise and (b) psophometric noise............................... 174
Fig. 5-1. Proposed interleaved single-stage/single-phase three-level phase-shift converter. ..... 180
Fig. 5-2. Main waveforms........................................................................................................... 181
Fig. 5-3. Operating stages: (a) first stage (to, t1), (b) first stage (t1, t2), (c) first stage (t2, t3), and
(d) first stage (t3, t4)..................................................................................................................... 184
Fig. 5-4. Simplified three-phase interleaved converter: (a) simplified interleaved single-stage TL-
PS converter, and (b) including AC capacitors to eliminate the neutral point connection of the
power system. ............................................................................................................................. 185
Fig. 5-5. Simulation results for Lr=2µH: (a) inductor currents, (b) interleaved current, and (c)
amplitude spectrum..................................................................................................................... 187
Fig. 5-6. Simulation results for Lr=5µH: (a) inductor currents, (b) interleaved current, and (c)
amplitude spectrum..................................................................................................................... 187
Fig. 5-7. Simplified interleaved circuit using an auxiliary transformer to compensate for a large
resonant inductance used to increase the load range with ZVS operation.................................. 189
Fig. 5-8. Primary simulated waveforms: (a) transformer current for both simplified interleaved
and non-interleaved single-stage TL-PS converter, and (b) voltage applied across points A and B
for both converters. ..................................................................................................................... 191
Fig. 5-9. Experimental results obtained at 3kW and 50kHz: (a) boost inductor and interleaved
currents at Vin=220V, (b) amplitude spectrum of the interleaved current at Vin=220V, and (c)
interleaved current measured at three different input voltages................................................... 193
Fig. 5-10. Simulated high-frequency harmonics of the interleaved current. .............................. 194
xiv
Fig. 5-11. Boost inductance vs. frequency: (a) 3kW TL-PS simplified interleaved converter and
(b) 3kW VIENNA rectifier. ........................................................................................................ 195
Fig. 5-12. Results of comparison: (a) filter inductance L1, (b) filter inductance L3, (c) filter core
weight, (d) combined filter core and winding weight per phase, (e) core weight of filter and boost
inductors per phase, and (f) overall filter and boost inductors weight per phase (core + winding
Cu)............................................................................................................................................... 196
xv
List of Tables
Table 2-1. Extending the results given in (2-14) to one quarter of the line period....................... 28
Table 2-2. Normalized average current through the output diode as a function of phase θ. ........ 31
Table 2-3. System parameters....................................................................................................... 37
Table 2-4. Mathematical representation of the high-frequency inductor currents at different
operating stages during the interval 0<θ<π/6 of the AC voltages. ............................................... 44
Table 2-5. System parameters....................................................................................................... 56
Table 2-6. Identifying the DM filter design point (example: fs=40kHz)...................................... 62
Table 2-7. System parameters and components designed for 6kW application. .......................... 76
Table 2-8. Benchmarking the single-switch DCM boost rectifier (Vin=220V LN)...................... 77
Table 3-1. Identifying the DM filter design point (example: fs=40kHz)...................................... 99
Table 3-2. Benchmarking the two-switch three-level DCM boost rectifier (Vin=220V LN) . ... 116
Table 4-1. Components used in the implementation. ................................................................. 156
Table 4-2. Benchmarking the single-stage front-end converters (Vin=220V LN)...................... 175
Table 5-1. Components used in both interleaved and non-interleaved single-stage TL-PS
converters.................................................................................................................................... 189
xvi
List of Acronyms
AC: alternating current
CC: Cauer-Chebyshev
CCM: continuous conduction mode
CM: common mode
DC: direct current
DCM: discontinuous conduction mode
DM: differential mode
DPS: distributed power system
EMI: electromagnetic interference
IEC: International Electrotechnical Committee
LISN: line impedance stabilization network
LN: line-to-neutral
NPC: neutral-point-clamped
PFC: power factor correction
PWM: pulse width modulation
RMS: root mean square
THD: total harmonic distortion
TL-AS: three-level asymmetrical
TL-PS: three-level phase-shift
ZCT: zero-current transition
ZVS: zero-voltage switching
ZVZCS: zero-voltage and zero-current switching
xvii
1. Converters for Distributed Power Systems
A distributed power system (DPS) offers many advantages from the standpoint of high power
capability, reliability, modularity, redundancy and maintainability [1]. As a result, DPS has
become rather popular in telecom and server applications. For centralized power supplies, the
reliability must be very high, since a failure would cause the entire system to shut down.
However, the failure of any power module in a DPS has a reduced effect on the overall system
because of the built-in N+1 redundancy, where N is the minimum number of modules needed to
supply the load. Additional advantages of DPS are rapid replacement of faulty modules and
flexibility to expand the system capacity as the load requirements increase [2].
Despite several important advantages, a DPS offers two major drawbacks: (1) extra cost and (2)
noise caused by several converters placed next to each other [3]. Thermal issues may arise
because the power supplies are either compressed into ever-smaller cases or because they are
mounted on the logic boards to which they supply power. Nevertheless, paralleling decreases the
dissipation per module because each module is required to handle less power in the system,
A typical block diagram of a DPS is illustrated in Fig. 1-1. As can be seen, a DPS consists of
several stages of power conversion. The name distributed power systems alludes to the fact that
the power processing units in the system may not be located in the same place, but are distributed
according to load type and location [1] [2]. For the structure shown in Fig. 1-1, the front-end
converter is supplied by an AC power source, which can be either a single- or a three-phase bus.
The power is processed by the DPS front-end converters represented by the power supplies (PSs)
DC/DC front-end converter used to regulate the 48V DC distributed power bus. Small DC/DC
high-density power modules are then distributed according to load requirements and location to
provide point-of-load regulation. Besides supplying power to the load, the front-end converter
for telecom applications, for instance, is also used to charge the backup batteries connected
across the distributed DC bus voltage (not shown in Fig. 1-1). As described in section 1.4, the
work described hereafter concentrates on the front-end converter of the DPS. The load converters
are not discussed in this dissertation, as they have already been approached by other authors [4]
[5].
Front-end converters for DPS applications are typically used in telecom and server applications.
There is a wide range of front-end converters available in the market for telecom applications,
with power levels ranging from hundreds of watts to several kilowatts. For server applications,
the typical power level is 1kW. However, the market for sever applications is growing quickly,
as is the power level required for such applications. As predicted in the past, computer
applications will continue to drive the power supply industry and to promote the widespread use
of DPSs [3].
The PFC stage of the front-end converter has become an important accessory because, especially
in Europe, several standards are now limiting the emission of harmonic currents caused by
electronic equipment [6]-[8]. However, it has always been difficult to justify the cost incurred by
adding the PFC stage to the DPS front-end converters, especially for high-power applications
(6kW and up). Consequently, the research on PFC circuits for high-power applications is still
open in terms of finding solutions able to significantly reduce cost. Obviously, the performance
PFC circuit. Reducing the cost of the PFC stage, while preserving the overall performance
indices of the DPS front-end converter is rather challenging. As a result, this issue needs to be
carefully addressed.
P S #2
DC/DC #2 Load
…
…
P S #N
DC/DC #M Load
P S #N+1
Input
PFC DC/DC
Filter
For the next generation of high-power DPS front-end converters, not only the overall
performance but also the cost of the entire system will be important issues to be considered
during the design process. The front-end converter for DPS applications must achieve high
power factor, low harmonic distortion, high efficiency, high power density, high reliability and
low electromagnetic interference (EMI) noise. To reduce the cost of the front-end converter, the
PFC stage must be inexpensive, while still complying with standards for harmonic distortion.
converters, is the use of a two-stage approach based upon single-phase power modules, as
discussed in previous papers [9]-[11] and shown in Fig. 1-2. That figure shows that the first stage
of each module is used to perform the PFC function to meet harmonic current standards such as
the IEC 61000-3-2, while the second-stage DC/DC converter regulates the DC output voltage of
The PFC circuit of each module operates in the continuous conduction mode (CCM) and is
controlled by average current mode control. The performance of the two-stage approach is high,
but so is the cost involved because of the number of components used to realize the system.
Moreover, to operate in CCM, the single-phase PFC circuit requires a complex control, as well
as a large intermediate bus capacitor to limit the bus voltage ripple and handle the pulsating
power in the intermediate DC-link bus. Nevertheless, the two-stage DPS using single-phase
front-end modules has the advantage of modularity, which is also a measure of performance,
A possible approach for reducing the cost of the front-end converter is to simplify the PFC stage.
Connecting the three outputs of the single-phase boost rectifiers eliminates the pulsating power
in the high-voltage DC-link intermediate bus, while permitting the use of only one DC/DC
converter, as shown in Fig. 1-3. In this way, the volume of the intermediate bus capacitor can be
reduced [12] because the pulsating power is eliminated. However, the direct connection of the
outputs of the single-phase boost rectifiers creates interactions between the PFC circuits.
Filter
EMI
48V
Bus
B
PWM - Dideal -
+
Phase-Shift Ho(s) Go(s)
Hi(s) +
Vref
-
+
- Vbus
kv x Gv(s) Current
+ Sharing Bus
B
Single-Phase Module
C
C
Single-Phase Module
A
Fig. 1-2. High-power DPS front-end converter using three single-phase modules.
Filter
EMI 48V
Bus
B
PWM -
- +
Phase-Shift Ho(s) Go(s)
Hi(s) +
Vref
-
+
- Vbus
kv x Gv(s)
+
B
Modified Boost
C
C
Modified Boost
A
Fig. 1-3. Simplified front-end converter using three single-phase rectifiers connected to the same output.
parts, while a diode is added to the circuit configuration of each single-phase boost rectifier. For
proper operation, the intermediate bus voltage must be at least twice the peak input voltage,
which means that the voltage rating of the boost power switches must be at least 800V in
applications for which the nominal line-to-line input voltage is 380V. Eventually, the switches of
the DC/DC converter will experience higher voltages stress unless a three-level structure is used
for the DC/DC converter [13]-[16]. Another approach used to reduce the cost of the DPS has
been reported [17]. However, that system has been implemented for 500W of total power, and a
three-phase line-to-line voltage of 200V has been selected in order to limit the voltage stress
across the Sepic switch to less than 500V. For higher-input-voltage applications, the Sepic
converter requires switches with much higher voltage rating, which makes the solution not viable
rectifiers [18] [19], as shown in Fig. 1-4(a). The VIENNA rectifier can be seen as a simplified
version of three single-phase PFCs connected to the same intermediate bus voltage, as illustrated
in Fig. 1-3. The main idea behind this simplification is to use the neutral-point connection of the
split bus capacitors to reduce the voltage applied across the power switches. For proper
operation, the bus voltage still needs to be at least twice the line-to-neutral peak input voltage.
However, the three-level structure obtained by using the neutral-point connection reduces the
voltage stress across the switches to half of the total bus voltage, thus allowing 500V MOSFETs
to be used in the VIENNA rectifier. Another version of the VIENNA rectifier was presented
reason it could not solve the problem of the voltage stress across the power switches.
DC link
(a)
Phase A Phase B Phase C
DC link
Io Vo
-
Vref
+
PWM
- Gdc(s)
Hi(s) Phase-Shift Hdc(s)
- +
+ + Vo/2
Fc(s)
- + -
Kv X Gv(s)
+ Vo
(b)
Fig. 1-4. Using the VIENNA rectifier in the PFC stage of the high-power front-end converter: (a)
VIENNA rectifier and (b) control block diagram and connection of DC/DC converter.
In the VIENNA rectifier, the switches of the DC/DC converter must resist the total bus voltage,
as illustrated in Fig. 1-4(b). There are two possible ways to overcome this problem: the
DC/DC converter can be used to reduce the voltage stress across the power switches [21].
Another important achievement of the VIENNA rectifier is that a phase-leg integrated power
module can be purchased to implement each leg. The integrated module is provided by IXYS as
part numbers VUM 25-05 and VUM 85-05. Both modules consist of the integration of the power
switch, the diodes of the single-phase bridge, and the two fast diodes (one connected to the
There are many other approaches that can be used to enhance power quality in high-power
applications. Among the three-phase rectifiers, the six-switch boost topology is able to achieve
the best performance in terms of shaping the input currents and presenting reverse energy flow
capability [22] [23] [24]. However, as shown in Fig. 1-5(a), it is necessary to use IGBTs instead
of MOSFETs because of the high bus voltage processed by the boost-type rectifier. High-voltage
MOSFETs add conduction loss to the circuit, thus reducing the efficiency. Additionally, when
MOSFETs are used in the circuit, the anti-parallel diodes will present serious reverse-recovery
problems due to the high operating frequency. To improve the efficiency of the six-switch boost
rectifiers, several soft-switching schemes and lossless snubbers have been presented [24] [25]
[26]. Either soft-switching circuits or lossless snubbers add cost to the system, making these
solutions difficult for industry to accept. Further improvements in switching and conduction
losses have also been achieved by applying special modulation schemes that optimize the
best device choice for six-switch boost rectifier applications, one has to be cautious when using
IGBTs because their turn-off loss limit the switching frequency to well below 40kHz, except if
soft-switching schemes are used to improve the turn-off conditions for the IGBTs.
(a)
(b)
Buck-derived rectifiers can also draw sinusoidal input currents from the mains [28] [29] [30]
[31] [32] [33]. The buck rectifier offers some advantages over the six-switch boost approach,
current, and low bus capacitance. On the other hand, the major disadvantages of the buck
rectifiers, as compared to the boost topologies, are the pulsating input currents and the
conduction loss caused by several voltage drops in the path of the inductor current. As shown in
Fig. 1-5(b), a freewheeling diode can be used to reduce the conduction loss in the buck rectifier.
However, this solution eliminates rectifier’s ability to handle bi-directional power flow.
The unidirectional PFC approach is a good option for reducing the cost of the front-end PFC
stage [18] [19]. Examples of unidirectional PFC approaches have been proposed [34] - [36], as
depicted in Fig. 1-6(a). The circuit under consideration is able to provide PFC over a wide input
voltage range. At high-line input voltage, the range switch must be open, while the range switch
must be closed at low-line input voltage in order to increase the rectifier voltage gain. The circuit
shown in Fig. 1-6(a) uses the upper switch to control the most positive input current and the
lower switch to control the most negative input current. Since the rectifier is able to control only
two input currents at a time, selecting one of the three low-frequency switches according to the
input voltages actually indirectly controls the third line current as well. The three low-frequency
switches can be easily realized by one of the configurations shown in Fig. 1-6(b). The low-
frequency switches do not experience switching stresses, and the RMS current is reduced
(a)
(b)
Fig. 1-6. Unidirectional PFC used for wide input voltage range: (a) topology with range switch and
(b) different ways of realizing the three low-frequency switches.
For very high-power applications, low-frequency PFC topologies have been used as low-cost
front-end PFC circuits. The use of low-frequency rectifiers, however, implies that size is not a
topology shown in Fig. 1-7(a). For this application, the three-level rectifier operates at low
frequency, and the power switches shape the input currents by conducting during short intervals
relatively low at heavy load, it drastically increases as the output power decreases [37].
(a)
Id+I3
Ias Iar
Id-I3
Iaj
2I3
(b)
Fig. 1-7. Low-frequency PFC solutions: (a) three-level PFC topology and (b) injecting third
harmonic.
[39], as illustrated in Fig. 1-7(b). The switches are operated at high frequency in order to control
the currents through the DC-side inductors. However, despite the high-frequency operation of the
power switches, the series network must be tuned around the third harmonic of the line
frequency to provide low THD for the input current. As a result, the size of the converter is large
Another option for very high-power installations is the use of either a shunt or a series active
filter, or a combination of both, to improve the quality of the input currents generated by non-
linear loads [40] [41], such as power supplies. This type of solution is centralized and does not
require local PFC capability for individual loads. Fig. 1-8(a) illustrates a shunt active filter used
to shape the input source current. The active shunt filter is not connected to the main path of the
input current. Therefore, the active shunt filter is required to process only part of the total load
power, as opposed to rectifiers that are always required to process 100% of the power. Two
simulation results are provided in Fig. 1-8 to demonstrate the extent to which the use of active
shunt filters is advantageous when compared to rectifiers. For both simulation results, the load
requires a total active power of 15kW. In Fig. 1-8(b), the current THD generated by the load is
30%, which requires the active filter to process only 3.6kVA. On the other hand, Fig. 1-8(c),
shows that the higher load current THD requires the active filter to process much more apparent
power. For this example, the active filter has to process 16.2kVA, which is even higher than the
active power required by the load. Therefore, these two simulation results show that the use of
active shunt filters is advantageous when the harmonic distortion generated by the load is not too
iload
Non Linear
N
Load
if
(a)
load
load
filter
50A/div filter
260 270 280 290 300 260 270 280 290 300
Time (ms) Time (ms)
(b) (c)
Fig. 1-8. Active shunt filter used to improve the current waveform generated by non-linear loads:
(a) structure, (b) 30% of THD produced by the load current, and (c) 100% of load current THD.
Because several comparisons will be made during the course of this dissertation, it is rather
important to have a baseline with which compare the results that will be presented hereafter. A
benchmark circuit should be based on some kind of practice widely adopted in industry. For the
PFC front-end stage, the state-of-the-art approaches for high-power DPS applications are based
upon the three single-phase modules and the VIENNA rectifier. Therefore, throughout the course
of this dissertation, the results that are obtained will be compared against the benchmark circuits
shown in Fig. 1-2 and Fig. 1-4. Several aspects are benchmarked, including size of the
differential mode (DM) input filter, size of the boost inductors, and device stress.
The front-end converters for DPS applications are extensively used in telecommunication
systems to supply load converters and to charge batteries to provide backup energy during power
grid blackouts. Besides telecom front-end converters, mainframe computers and server
applications also require front-end converters to distribute power to the loads. Although the
typical power level for server applications is 1kW, it is envisioned that in the future high-end
servers will demand more power, which justifies the development of front-end converters for
There are two major reasons for developing research on PFC circuits used for high-power DPS
applications. The first reason is to improve power quality, since the harmonic current must be
limited within the strict bounds established by standards. Otherwise, the manufacturer of front-
end converters would not be able to promote a competitive product in the market. The second
reason for researching PFC is cost reduction. Although customers require PFC, they are not
Based on the previous discussion, this dissertation presents an effort towards investigating
simple and low-cost solutions for the three-phase PFC used in DPS applications. The main
purpose of this dissertation is to devise PFC techniques and circuits for 3kW to 6kW applications
that achieve the following features: (1) reduced complexity and (2) reduced cost, while
To realize these objectives, chapter 2 discusses the drawbacks and presents improvements to the
DCM boost rectifiers is derived to achieve input current ripple cancellation, and consequently to
reduce filtering requirements. Chapter 3 explores a two-switch three-level DCM boost rectifier
with improved performance. Because of the voltage stress reduction brought about by using a
three-level topology, one can use MOSFETs with low Rds-on to simultaneously increase the
switching frequency and efficiency of the PFC circuit. Combining the interleaving technique to
cancel the input current ripple with increased switching frequency further reduces input filtering
In the first part of the dissertation, two-stage approaches are the main focus of the work. Both the
single- and two-switch PFC circuits developed in chapters 2 and 3 are interfaced with a high-
power three-level DC/DC converter to evaluate the performance of a two-stage approach. All the
The second part of this dissertation explores single-stage converters for mainframe computers
and servers applications. Two topologies are presented, one of which is further developed to
achieve simplified interleaving, thus avoiding the duplication of the entire switching power
stage. Chapter 4 presents two novel three-phase single-stage front-end converters implemented
for 3kW applications. The motivation for developing single-stage converters is the potential that
these approaches present for achieving cost reduction. Both topologies presented in chapter 4 are
based on the functional integration of the two-stage approach presented in chapter 3. Three-level
topologies are suitable for this type of application because of the voltage stress reduction across
the power switches. The first single-stage converter presented in chapter 4 is based on the three-
level phase-shift converter, while the second single-stage approach employs an asymmetrical
PWM converter to transfer power and to regulate the DC output voltage. Experimental results
and comparisons are presented throughout the dissertation to clarify drawbacks and advantages
of both single-stage converters, as well as to verify how they stand up against the benchmark
Chapter 5 is devoted to developing an interleaved single-stage converter that eliminates the need
for duplicating the entire switching power stage in order to provide input current ripple
cancellation. The features of this technique are analyzed, and experimental results are obtained
for a 3kW prototype. Chapter 6 wraps up all the results and summarizes the conclusions drawn
2.1. Introduction
Different PFC circuits for DPS applications have been proposed in recent years. For higher
power levels (6kW and higher), the need for three-phase rectifiers is clear. However, one of the
key practices used by industry to obtain high-power front-end converters is the connection of
single-phase power modules to the three-phase AC system, as discussed in section 1.1, and
shown schematically in Fig. 2-1(a) [9]-[11]. For a more detailed building block configuration of
To reduce the cost of the rectifier system, Fig. 2-1(b) shows three single-phase CCM boost
rectifiers connected to the same intermediate bus voltage [12]. In this case, the system cost is
reduced because only one DC/DC converter is used to regulate the DC output voltage, as
opposed to the three DC/DC converters required in Fig. 2-1(a). Nevertheless, connecting the
outputs of the rectifiers creates undesirable interactions between the converters. To reduce such
interactions, an extra diode is added to each single-phase boost rectifier, while the boost inductor
is split into two inductors, as shown in Fig. 1-3. The voltage of the intermediate DC bus must be
at least two times higher than the peak line-to-neutral input voltage in order to guarantee proper
operation of the CCM boost rectifiers. Although the system shown in Fig. 2-1(b) is simplified
with respect to the system illustrated in Fig. 2-1(a), the modularity is lost because the CCM boost
While both approaches for PFC result in high performance, they are high cost because of the
circuit complexity and number of components. Therefore, it is clear that lower cost three-phase
pros and cons of using discontinuous conduction mode (DCM) boost rectifies for telecom and
(a)
Input CCM
B
filter PFC
Input CCM
C
filter PFC
(b)
Fig. 2-1. General approach used to obtain PFC for high-power-level applications: (a) single-phase
modules and (b) simplified approach.
One of the approaches that can be used to reduce the cost and complexity of the DPS front-end
converter for high-power applications is the single-switch three-phase DCM boost rectifier
shown in Fig. 2-2 [42]. This topology offers the advantages of simplicity and low harmonic
distortion. The following sections are devoted to demonstrating the approaches used to improve
The main contributions of this chapter are the analysis of the interleaved input current ripple
(section 2.5), the analysis of the best switching frequency range to reduce the combined filter and
converter (section 2.8). The analysis of the operation shown in section 2.2 has already been
described [43] - [45], and repeated hereafter for the purpose of background information.
EMI
filter
PWM -
Gv(s)
+
Vbus
This section provides a simplified analysis and design for the single-switch DCM boost rectifier.
Because of the symmetry of three-phase systems, the analysis of the operating stages can be
limited to the interval 0<θ<π/6. For the purpose of mathematical representation of the DCM
va = V pk sin (θ )
2π
(2-1) vb = V pk sin θ − ,
3
2π
vc = V pk sin θ +
3
rectifier, as shown in Fig. 2-3. In the same figure, k represents the kth switching period within
the line period Tr. In the first operating stage, the power switch is turned on to linearly charge the
input inductors according to the phase voltage that is applied across each one. In the second
operating stage, the power switch is turned off to reset the inductors. The inductor with the
lowest peak current resets first. In the third operating stage, the two remaining inductor currents
are reset to zero at the same rate. Once the reset interval has finished, the output load is supplied
by the energy stored in the output filter capacitor until the next switching period restart. In the
analysis that follows, the line-to-neutral input voltage va is taken as the reference voltage for the
Ts
Icp
Ics
ic
Iap
ia
ib
Ibs
Ibp
(k-1)/fs ton tr ts td
Va + Va +
~ ~
Vb + Vb + Vo
~ + ~
Vc + Vc +
~ ~
1st stage - ton 2nd stage - tr
Vb + Vo
~ +
Vc +
~
3rd stage - ts 4th stage - td
Fig. 2-3. Current in the boost inductors and operating stages in the interval 0<θ<π/6 of the input
AC voltages.
The switching frequency of the converter is much higher than the line frequency. Therefore, the
line voltages can be considered constant within one switching period. In this case, the peak line
va D
I ap =
L fs
vb D
(2-2) I bp = ,
L fs
vc D
I cp =
L fs
where the quantities above Iap, Ibp and Icp are the peak line currents at the end of the on-time
interval, D is the duty cycle, L is the input boost inductance, fs is the switching frequency, and ton
To determine the characteristics of the DCM boost rectifier, it is necessary to know the
instantaneous average value of the line currents as a function of θ. Once the average line currents
have been described within the interval 0<θ<π/6, it is then possible to extend the results to the
entire line period of the input voltges by using the symmetry properties of three-phase systems.
Taking into account the operating stages and the time diagram shown in Fig. 2-3, one can write
the expressions of the average line currents for one switching cycle as a fuction of the phase
angle θ, as follows:
(2-3) ib =
t on I bp (θ )
+
[I (θ ) + I (θ )]t
bs bp r
+
t s I bs (θ )
,
2 Ts 2 Ts 2 Ts
ic =
t on I cp (θ )
+
[I (θ ) + I (θ )]t
cs cp r
+
t s I cs (θ )
2 Ts 2 Ts 2 Ts
where ia, ib and ic are the average line currents as a function of θ, while Ibs and Ics are the currents
through lines b and c at the end of the second stage, tr is the time duration of the second
operating stage, ts is the time duration of the third operating stage, and Ts is the switching period.
The duration of the second stage is the time taken by the current through inductor La to be reset:
L (0 − I ap )
(2-4) tr = .
v La
The voltage vLa across the inductor connected to line a can be obtained from the equivalent
network of the second operating stage, as shown in Fig. 2-3. As a result, the second operating
v a − v La − Vo + v Lb − vb = 0
(2-5) v a − v La + v Lc − vc = 0 .
v La + v Lb + v Lc = 0
Solving the above expressions for the voltage across the input inductors during the second stage
Substituting (2-2) and (2-6) into (2-4), results in the time duration of the second operating stage
as a function of output and input voltages, duty cycle and switching frequency:
D 3 va
(2-7) tr = .
f s (Vo − 3 v a )
During the second operating stage, the voltage across the input inductors connected to lines b and
L (I bs − I bp )
v Lb =
tr
L (I cs − I cp )
(2-8) .
v Lc =
tr
Equations (2-2), (2-6) and (2-7) can be substituted in (2-8). The resulting set of equations can be
solved for the current through lines b and c at the end of the second operating stage:
D (vb + 2 va )Vo
I bs =
L f s Vo − 3 v a
(2-9) .
D (vc − v a )Vo
I cs =
L f s Vo − 3 va
calculate the average current through the power lines. From Fig. 2-3, the equivalent network of
vb − v Lb + Vo + v Lc − vc = 0
(2-10) .
v Lb + v Lc = 0
In order to determine the duration of this stage, it is necessary to write down the equation of the
L (0 − I bs )
(2-11) v Lb = −v Lc = .
ts
From (2-9), (2-10) and (2-11), one can obtain the expression of the time duration for the third
D − 2Vo (2 v a + vb )
ts = .
(2-12)
fs ( )
V o − 3 va (Vo + va + 2 vb )
From the previous analysis, the average currents through the power lines can be obtained by
substituting (2-1), (2-2), (2-7), (2-9) and (2-12) into (2-3). The result is described as follows as a
Equation (2-13) is not normalized, and for this reason it restricts the value of the analysis. The
D2 M sin (θ )
ian (θ ) =
2 (M − 3 sin (θ ))
π
2 3 sin (2θ ) − M sin θ +
D M 3
ibn (θ ) =
(2-14)
(
2 (M − 3 sin (θ )) M − 3 cos(θ )
,
)
3 π
−
2
sin (2θ ) − M sin θ −
D M 2 3
icn (θ ) =
(
2 (M − 3 sin (θ )) M − 3 cos(θ )
)
where:
Vo
M =
V pk
(2-15) .
ia ib ic L f s
ian ibn icn =
V pk
boost rectifier, while ian, ibn and icn are the normalized instantaneous average currents through the
The analysis performed in the first 30o interval of the input line voltages can be extended to the
entire line period. Because of the symmetry of the three-phase system, it is enough to expand the
analysis of the converter up to one quarter of the line period, as shown in Table 2-1. By phase-
shifting the expressions given in (2-14), one can extrapolate the average line currents to the
Table 2-1. Extending the results given in (2-14) to one quarter of the line period.
It is important to know the boundary between the continuous and discontinuous operating modes
so that the converter is designed to operate in DCM under all possible operating conditions. In
DCM operation, the currents in the boost inductors are zero before the power switch is turned on
again. The total time for which there is current circulating in the input lines can be determined
Substituting equations (2-1), (2-7) and (2-12) in the expression above results in:
D Vo 1
(2-17) ∆t = .
fs Vo − 3 V pk cos(θ )
When the converter operates in critical conduction mode, the time interval described above
reaches its maximum value. To determine the phase angle θ at which the time interval in (2-17)
reaches its maximum, the derivative of that expression must be taken with respect to θ, as
follows:
d∆t D Vo − 3 V sin (θ )
(2-18) = pk
.
dθ fs
o(
V + 3 V cos(θ ) 2
pk )
The time interval in (2-17) reaches its maximum when the derivative in (2-18) equals zero,
which happens at θ=0. Therefore, the maximum time interval given by (2-17) is:
D Vo 1
(2-19) ∆t max = .
fs Vo − 3 V pk
It is clear that the ∆tmax cannot be greater than the switching period Ts of the boost rectifier. As a
3
(2-20) M cr = ,
1− D
in the boundary between the continuous and discontinuous modes. For a given duty cycle, the
The output characteristics are a set of curves used to design the DCM boost rectifier, as they
represent the voltage conversion ratio as a function of the normalized average output current. To
plot the output characteristics, it is necessary to obtain first the output average current. From the
topology of the single-switch boost rectifier shown in Fig. 2-2, it can be verified that the average
output current equals the average current through the output diode. In the first 30o interval of the
line period that has been previously analyzed, the average current through the power rectifier, as
(2-21) id =
(I bs + I bp )t r
+
t s I bs
.
2T 2T
Substituting (2-1), (2-2), (2-7), (2-9) and (2-12) into (2-21) yields:
The normalized average current through the output diode as a function of the phase angle θ is
then given by
current through the output diode is six times greater than the line frequency. Therefore, in order
to describe a complete period of the average current through the output diode, the analysis only
Table 2-2. Normalized average current through the output diode as a function of phase θ.
solved:
π π
3 6
( )
3
idn (θ ) dθ + ∫ idn π − θ dθ .
π ∫0
(2-24) I on =
π
3
6
The solution of (2-24) is plotted in Fig. 2-4, and represents the output characteristics. As can be
observed, the voltage conversion ratio M is plotted as a function of the normalized average
output current Ion, using the duty cycle D as the running parameter. The curve that limits the
output characteristics is the boundary between the continuous and discontinuous modes. The
converter must be designed to operate inside the discontinuous mode of operation to guarantee
low harmonic distortion of the input current. As can be seen, the curves are rather steep, which
gives a current source characteristic to the single-switch three-phase DCM boost rectifier. His
characteristic enables parallel operation without the need to impose current sharing.
Voltage gain
2.5
D=0.1
2 D=0.25
D=0.4
DCM/CCM boundary
1.5
0 0.02 0.04 0.06 0.08 0.1 0.12
Normalized output current
Fig. 2-4. Output characteristics of the single-switch three-phase DCM boost rectifier.
This section describes the design of the DCM boost rectifier. In this example, the RMS line-to-
neutral input voltage ranges from 187V to 244V, the DC output voltage is 800V, the switching
frequency is 40kHz, the output power is 4kW, and the minimum power is 600W (the minimum
power could be lower if necessary). From these specifications and from (2-15), the voltage gain
M of the DCM boost rectifier is 3.03 at low-line input voltage and 2.32 at high-line. As shown in
Fig. 2-5(a), the rectifier is designed at high-line input voltage, and the design point is chosen near
the DCM/CCM boundary. This procedure optimizes the design of the DCM boost rectifier by
reducing the current stress in the components of the circuit. From the design point shown in Fig.
2-5(a), one can read the output normalized current as 0.065, which is combined with (2-15) and
given specifications to result in a boost inductance of 112µH. For this particular design, the
highlighted area inside the output characteristics seen in Fig. 2-5(a) defines the operating region
keep the boost rectifiers running in DCM. The duty cycle variation as a function of the
The IEC 61000-3-2 Class A standard limits the emission of harmonics generated by electronic
equipment with phase current of up to 16A and 230V of input line-to-neutral voltage [6]. For
higher-power equipment with phase currents higher than 16A, the recommendations established
in the IEC 61000-3-4 should be used instead [7]. Throughout this work, reference is made only
to the IEC 61000-3-2 harmonic standard because the line current is lower than 16A.
In the single-switch three-phase DCM boost rectifier, the fifth harmonic of the input current is
the dominant low-frequency harmonic, which is responsible for limiting the power that can be
extracted from the rectifier shown in Fig. 2-2, while still meeting the limits of the IEC standard.
Because of the input current harmonic distortion, the circuit shown in Fig. 2-2 cannot comply
with the IEC 61000-3-2 Class A harmonic standard at higher power levels (>6kW at 800V of bus
voltage and 220V of line-to-neutral input voltage). To comply with the IEC standard at several
kilowatts, the bus voltage of the DCM boost rectifier Vbus should be increased. This is not a
desirable solution because it will increase the voltage stress across the power devices of the
rectifier itself, as well as across the devices of the converter that will be connected across the
D=0.40 0.3
D=0.25
Voltage gain, M
Duty Cycle
3 0.25
0.35
0.3
0.25
Duty Cycle
0.1
0.05
2.3 2.43 2.57 2.7 2.83 2.97 3.1
Voltage gain, M
(c)
Fig. 2-5. (a) Operating region of the specified design, (b) duty cycle operating region versus
normalized output current, and (c) duty cycle operating region versus voltage gain.
To alleviate this problem to some extent, different modulation techniques have been proposed to
reduce the harmonic distortion of the input currents without increasing the bus voltage beyond
practical levels. The first approach proposed to improve the harmonic distortion of the input
currents involved operating the single-switch boost rectifier in the critical mode [44] [45]. To do
this, the power switch must be turned on at the instant at which the boost diode current reaches
zero. As a result, the switching frequency is variable, and the effective duty cycle modulation
over the line cycle results in reduced THD of the input currents. The drawback of operating the
Another approach for improving the THD of the input current involves controlling to a constant
level the average current in the boost diode [46]. In order to keep the average current constant
through the boost diode, the duty cycle must be modulated over the line cycle, resulting in an
improved input current waveform. The drawback of this method is the extra current sensor
A simple technique that can be used to reduce the harmonic distortion of the input current is the
so-called harmonic injection method [47]. This method is illustrated in Fig. 2-6(a), and the
principles for achieving optimal harmonic injection are described in other work [47]. Fig. 2-6(a)
shows that a modulating signal is added to the control signal in order to modulate the duty cycle
over the line period. The modulating signal is a sixth-order harmonic signal using an appropriate
phase angle extracted from the three-phase input voltages. By controlling the modulation index,
it is possible to improve the THD of the input currents. Fig. 2-6(b) shows the effect of the
harmonic injection on the spectrum of input currents. As can be observed, when the DCM boost
rectifier operates at 8kW of output power and with a constant duty cycle over the line period, the
fifth-order harmonic is well above the limit specified by the IEC 61000-3-2 Class A standard.
However, under the same operating conditions, the harmonic injection method is able to lower
the amplitude of the fifth-order harmonic to slightly below the limit, while still maintaining the
same bus voltage. Fig. 2-6(c) shows how the harmonic injection technique helps the boost
rectifier comply with the IEC standard at higher power levels. For 800V bus voltage and constant
duty cycle control, the maximum power that can be extracted from the DCM boost rectifier is
voltages, the hamonic injection technique increases the power limit to more than 8kW, while still
complying with the IEC standard. Therefore, the harmonic injection helps to increase the power
level that can be extracted from the DCM boost rectifier without exceeding the limits of
harmonic emissions established by the IEC 61000-3-2 Class A standard, while avoiding
EMI
filter
Three-phase
PWM
bridge rectifier
+ -
-1 Gv(s)
+ +
Vbus
(a)
1.8
Constant duty cycle 12
1.6 Meeting IEC 61000-3-2 A
Harmonic current (A)
Fig. 2-6. Three-phase DCM boost rectifier: (a) harmonic injection method, (b) harmonic currents at
220V (LN input voltage), 800V of bus voltage and 8kW of output power, and (c) benefit of harmonic
injection.
Due to the DCM operation of the power stage, the current ripple in the input boost inductors is
quite large, as compared to those of CCM rectifiers. A higher input current ripple requires a
To show the impact of the DCM operation on the amplitude of the high-frequency input current
ripple, Fig. 2-7 compares the high-frequency spectrum of the input currents for three different
cases: (1) single-phase CCM boost rectifier; (2) VIENNA rectifier; and (3) single-switch three-
phase DCM boost rectifier. Cases (1) and (2) refer to the benchmark circuits described in chapter
1 as the baselines for comparison throughout this dissertation. Table 2-3 summarizes the system
parameters used in the comparison. As can be verified, all cases draw the same power per phase
at the same switching frequency. The boost inductances of the CCM rectifiers (single-phase
CCM boost and VIENNA) were designed to limit the maximum peak-to-peak input current
ripple to below 25% of the fundamental input peak current value. The bus voltage in all cases is
chosen accordingly to allow proper operation. The system parameters used for the DCM boost
rectifier follow the design developed in section 2.2.4, except that the power is twice as high and
for that reason, the boost inductance is half of that described in section 2.2.4.
Vin (V)
System Power (kW) L (µH) Vbus (V) fs (kHz) Line-to-
Neutral
2.67 (Need
Single-Phase
three
CCM Boost 496 400 40 187
rectifiers to
Rectifier
supply 8kW)
VIENNA
8 496 800 40 187
Rectifier
Single-Switch
8 56 800 40 187
DCM Rectifier
VIENNA rectifiers are approximately the same. For both cases, the amplitude of the first high-
frequency harmonic is 1.1A. However, one can see that the DCM boost rectifier generates more
than 10A at 40kHz. Therefore, if the EMI filter were to be designed to limit the emission of high-
frequency noise according to the VDE 0871 Class B standard, the size of the input filter needed
for the DCM boost rectifier would be much larger than the filter size required to attenuate the
noise generated by the CCM rectifiers. To overcome this drawback, an alternative solution for
reducing the amplitude of the high-frequency spectrum generated by the DCM rectifier is to
interleave the operation of two or more converters, as addressed in the next section.
The EMI noise generated by the input current ripple of the DCM boost rectifier must comply
with standards that limit high-frequency conducted noise emissions in electronic equipment. In
order to attenuate and limit the ripple of the input current to levels below the specifications of
any given EMI standard regulation, an EMI filter must be used at the input of the DCM boost
rectifier.
The fundamental component of the high-frequency spectrum of the input current ripple for
interleaved rectifiers is centered on the switching frequency, more specifically at the side-band
frequencies fs±fr (fs is the switching frequency and fr is the line frequency). Therefore, the cutoff
frequency of the EMI filter must be chosen to be well below the switching frequency of the
DCM boost rectifier in order to provide appropriate noise attenuation. As a result, the EMI filter
connected at the input of the DCM boost rectifier becomes large, since its cutoff frequency is set
well below the switching frequency, and the ripple to attenuate is also high.
10-2
10-3
10-4
10-5
10-6
10-7
10 102 103
Frequency (kHz)
(b)
102
10
1
Amplitude (A)
10-1
10-2
10-3
10-4
10-5
10-6
10 102 103
Frequency (kHz)
(c)
Fig. 2-7. High frequency spectrum of the input current ripple: (a) single-phase CCM boost rectifier,
(b) VIENNA rectifier, and (c) single-switch three-phase DCM boost rectifier.
generated by DCM boost rectifiers. For instance, the critical mode of operation can also be used
to reduce the high-frequency amplitude spectrum of the input current ripple [44]. In critical
conduction mode, the frequency is modulated according to the amplitude of the AC input
voltages and output load. As a consequence of the frequency modulation, the fundamental
component of the input current ripple is no longer centered at a single frequency, but is instead
spread across the minimum and maximum values of the modulated switching frequency [48].
Therefore, the high-frequency amplitude spectrum of the input current ripple is reduced, which
in turn will reduce the size of the EMI input filter. Frequency modulation is beneficial in
reducing the amplitude spectrum of the input current ripple, but this technique complicates the
design of magnetic devices, increases the switching losses at high switching frequencies, and
requires auxiliary circuitry to keep the converter running in critical mode. To simplify this
approach while reducing the input current ripple, one can interleave the operation of two or more
constant-frequency DCM boost rectifiers [49]-[51]. Interleaved rectifiers are able to reduce the
amplitude of the input current ripple and increase the effective ripple frequency according to the
number of interleaved channels. A drawback of interleaving the converters is that the number of
components used in the system must be increased according to the number of interleaved
channels. Therefore, in order to maintain the low cost, the number of interleaved channels cannot
be high. Although the interleaving technique increases the number of components, the device
current ratings are reduced by a factor that depends upon the number of interleaved converters.
Fig. 2-8 shows the schematic representation of a two-channel interleaved system. Diodes D10 and
D20 have been added to the power stage in order to eliminate electrical interactions between the
rectifiers. The drawback of adding diodes D10 and D20 is that the gate signals require isolation.
shifting the gate signals also phase-shifts the input inductor currents connected to the same
phase. Therefore, the effective line current, which is the sum of the inductor currents connected
to the same phase, will present a lower input current ripple. Additionally, the dominant high-
frequency harmonic of the interleaved input current is not centered around fs, but around nfs,
As a result of the reduced input current ripple and incremental increase in the effective ripple
frequency, the EMI filter can be designed with a higher cutoff frequency and lower attenuation,
as compared to the case using a single rectifier. Consequently, the size of the DM input filter is
also reduced. In addition to reducing the size of the EMI input filter, the interleaving of boost
converters can also increase the overall power level of the rectifier system by sharing the total
power between the interleaved rectifiers. Despite these advantages, the interleaving technique
cannot improve the low-frequency harmonic content characteristic of the boost converters
This section presents the analysis of the interleaved input current ripple using a frequency
domain technique. The results of this analysis can be used to design the DM input filter [52]. Fig.
2-3 shows the high-frequency boost inductor currents for one switching period, sampled in the
interval 0<θ<π/6 of the AC input voltages. According to the operating stages of the DCM boost
rectifiers, and the timing diagram for one switching period as illustrated in Fig. 2-3, it is possible
iL1 D1 D2 D3
D8
iTotal
L1
~ EMI S1
D7
L2 C R
~ Input
L3
~ Filter
D4 D5 D6
D10
L6
π
(2-25) θ = (k − 1) 6 , k = 1, 2,L, p ,
p −1
where p is the number of high-frequency triangles that can fit into the interval 0<θ<π/6 of the
AC input voltages, and k represents the kth triangle sampled in that interval.
The boost inductor current is periodical, and for this reason it can be represented by the Fourier
series, as follows:
∞ 1 Tr
(2-26) ia (t ) = C o + 2 ∑ ∫ ia (t ) e
− jω t
dt cos(ω t − θ m ) ,
m =1 Tr 0
As mentioned above, it is assumed that an integer number of high-frequency triangles can fit into
Ts 2Ts
∫ i∆ (t ) e − j ω t dt + ∫ i∆ (t ) e − j ω t dt +
1 0
1 2
∞
i (t ) = C o + 2 ∑ cos(ω t − θ m ),
Ts
(2-27) v = 1,2,...,12 p .
m =1 Tr v Ts
... + ∫ i∆v (t ) e
− jω t
dt
(v −1)Ts
Tr
=12 p
∞ Ts
(2-28) i (t ) = C o + ∑ 2 f r ∑ Fv ( j ω ) cos(ω t − θ m ) ,
m =1 v =1
where 12p is the total number of high-frequency current triangles that can fit into one period of
v Ts
(2-29) Fv ( j ω ) = ∫ i∆v (t ) e − j ω t dt .
(v −1)Ts
The current i∆v(t) in (2-29) is the high-frequency boost inductor current for every switching cycle
within the line period. From (2-28), the amplitude spectrum of the boost inductor current for one
D 3D Va 2 Vb (Vo + Vb + Vc )
Time
t on = tr = ts = −
fs f s (Vo − 2 Va + Vb + Vc ) f s (Vo − 2 Va + Vb + Vc ) ⋅
(V + V − V )
o b c
For n interleaved rectifiers, the gate signals are phase-shifted with respect to each other by Ts/n.
Therefore, the amplitude spectrum of the input interleaved current for n interleaved boost
rectifiers is given by
The results of the previous analysis can be implemented on a computer algorithm used to
calculate the amplitude spectrum of the interleaved input current for any number of interleaved
Besides determining the spectrum composition, one must also determine the operating point
related to the worst-case ripple condition in order to design the input DM filter parameters. The
the dominant high-frequency harmonic of the interleaved input current usually dictates the
design of the input DM filter2. For n interleaved DCM boost rectifiers, the dominant high-
frequency harmonic of the interleaved current occurs at the side-band frequencies (nfs ± fr), which
Fig. 2-9(a) shows the maximum normalized amplitude of the dominant high-frequency harmonic
of the input current for two interleaved DCM boost rectifiers as a function of the voltage gain M.
There are two curves represented in Fig. 2-9 for each graph: one is plotted for the design related
to the 187V to 244V line-to-neutral input voltage variation, while the other curve is related to the
170V to 265V voltage range. The latter voltage range is more common for three-phase
applications. Fig. 2-9(a) was obtained from the previous analysis by fixing the voltage gain and
2
This affirmation is true when the VDE 0871 EMI standard is used to design the input filter, since it starts limiting
noise at 10kHz. For other standards, such as the EN 55022 (CISPR 22), the situation requires further analysis since
the initial frequency limiting noise is 150kHz.
shown in Fig. 2-5(c). By changing the duty cycle from minimum to maximum and fixing the
voltage gain, the maximum normalized high-frequency harmonic of the interleaved current can
be identified and plotted as shown in Fig. 2-9(a). In a similar fashion, Fig. 2-9(b) represents the
operating duty cycle associated with the maximum amplitude of the dominant high-frequency
0.05
Maximum normalized high-frequency
0.04
0.035
0.03
0.025
2 2.2 2.4 2.6 2.8 3 3.2 3.4
Voltage gain (Vo/Vpk)
(a)
0.3
Duty cycle at maximum high-frequency
0.25
harmonic
0.23
0.2
0.18
0.15
2 2.2 2.4 2.6 2.8 3 3.2 3.4
Voltage gain (Vo/Vpk)
(b)
Fig. 2-9. Results of the interleaved current analysis: (a) maximum normalized amplitude of the
dominant high-frequency harmonic for two interleaved DCM boost rectifiers and (b) duty cycle
associated with the maximum dominant high-frequency harmonic.
voltage, which is represented by the highest voltage gain. As a result, the DM input filter must be
designed at low-line input voltage to guarantee the attenuation of the worst-case interleaved
ripple condition. The break point shown in Fig. 2-9(b) occurs because for a specific voltage gain
there are two different duty cycles associated with the same maximum amplitude of the dominant
high-frequency harmonic. After determining the operating point that leads to this maximum, the
results of the previous analysis can be used to calculate the parameters of the DM input filter at
For the design example given in section 2.2.4 (187V to 244V of input voltage variation), the
voltage gain at low-line input voltage is 3.03. Taking into account this voltage gain in Fig.
2-9(a), the maximum normalized high-frequency dominant harmonic equals 0.044. From (2-15)
and section 2.2.4, the denormalized high-frequency harmonic of the interleaved current results in
2.625A. The duty cycle at which the maximum dominant high-frequency harmonic occurs is
Fig. 2-10 shows the results obtained when the interleaved system works at the operating point
identified in the previous paragraph. Fig. 2-10(a) shows the calculated low-frequency spectrum
of the interleaved current, as well as the spectrum of the current produced by a non-interleaved
DCM boost rectifier operated at the same power level as the interleaved system. As seen in the
figure, the interleaving technique has no effect on the cancellation of low-frequency harmonics.
For the same operating point, Fig. 2-10(b) shows the calculated high-frequency spectra of the
interleaved and non-interleaved systems at the same output power level. As can be observed, the
interleaving technique cancels out the odd high-frequency harmonics of the interleaved input
60Hz, whereas for the non-interleaved DCM boost rectifier it occurs at 40kHz ± 60Hz. Fig.
2-10(c) shows the measured input current for the interleaved system operated at low line (187V)
and D=0.235 (worst-case interleaved input current ripple condition). To validate the analysis
presented in this section, Fig. 2-10(d) compares the experimental and theoretical frequency
spectra of the interleaved current at the worst-case ripple condition. The results show very close
agreement between theoretical and experimental results, which validates the analysis derived
above that is used to predict the amplitude spectrum of the interleaved input current.
A two-channel interleaved DCM boost rectifier has been implemented for 8kW, as shown in Fig.
2-11(a). Ferrite E55 cores were used to implement the 112µH boost inductors, IXSN35N1200U1
IGBTs for the main switches, DSEI 30-10A fast diodes for the rectifiers, and IXTN15N100
MOSFETs for the auxiliary zero-current transition (ZCT) switches. The resonant components of
the ZCT circuit were Lr=5.7µH (E21 ferrite core) and Cr=40nF/630V (polypropylene capacitor).
The design procedure of the ZCT circuit has been described previously [53]. Fig. 2-11(b) and
Fig. 2-11(c) show the results obtained from the interleaved system operated at 8kW (4kW per
channel). The two upper traces represent the input voltage and filtered input current, while the
lower trace is the interleaved current. As can be observed, two-channel interleaved boost
rectifiers are quite effective in providing input current ripple cancellation. The effectiveness of
reducing the size of the input filter by interleaving the operation of two rectifiers will be
Non-interleaved
10-1 10-1
Non-interleaved
boost rec. (2.8kW)
10-2 10-2
Interleaved boost
rectifiers (2.8kW)
10-3 10-3
10 102 103 104 10 102 103 104
Frequency (Hz) Frequency (kHz)
(a) (b)
10
10-1
Amplitude (A)
10-2
10-3
10-4
Experimental
10-5
Theoretical
10-6
104 105 106
(c) Frequency (Hz)
(d)
Fig. 2-10. (a) Low-frequency spectrum of the interleaved current at 187V, D=0.235, and Po=2.8kW
total, (b) high-frequency spectrum under the same operating condition, (c) experimental
interleaved input current at worst-case ripple (5A/div), and (d) comparison between experimental
and theoretical spectra.
The harmonics of the input current obtained at 8kW with constant duty cycle modulation do not
comply with the IEC 61000-3-2 standard. However, the harmonic injection method is able to
improve the THD and help the interleaved system meet the standard. The load sharing between
the two interleaved rectifiers is shown in Fig. 2-12(a). The two converters naturally share the
load, without requiring current mode control. This result is a consequence of the fact that the
DCM boost rectifier has a current source characteristic, which facilitates natural current sharing.
The first design is the same one discussed in section 2.2.4, while the second design was targeted
at 6kW of total power using an input line-to-neutral voltage variation of 170V to 265V. At 220V
line-to-neutral voltage, the efficiency of the first design was 94.5%, while the second design
resulted in 96%. The second design used the same devices, except that the boost inductance was
set to 140µH.
L1
~ EMI L2
~ Input
L3
~ Filter
L4
L5
L6
(a)
Vin Vin
iin iin
iL1+iL4 iL1+iL4
(b) (c)
Fig. 2-11. Two-channel interleaved DCM boost rectifiers: (a) configuration, (b) results without
harmonic injection at Vin=220V (line-to-neutral), Po=8kW, fs=40kHz, and Vo=800V (THD=12.7%), and
(c) results with harmonic injection (THD=10.8%). All current traces are 20A/div and all voltage
traces are 200V/div.
98
Design for 8kW
97
Design for 6kW
Efficiency (%)
96
95
94
93
92
160 180 200 220 240 260
Line-to-neutral RMS input voltage (V)
(b)
This section compares three different control strategies and their effect on the low- and high-
frequency spectra of the input current of DCM boost rectifiers [54]. For the purpose of
comparison, the operating point of the single-switch DCM boost rectifier is Vin=220V, Vo=800V,
single-switch DCM boost rectifier, are (1) fixed switching frequency with fixed duty cycle over
the line period, (2) fixed frequency with harmonic injection method, and (3) variable switching
frequency.
Fig. 2-13(a) shows the low-frequency spectrum of the input current for all the control strategies
at the operating point defined above. As can be observed, all control strategies produce the same
harmonic current at the line frequency, which guarantees that the same power is being delivered
to the output in all cases. The fixed duty cycle control produces the highest fifth-order harmonic
among all control strategies, which is a major factor limiting the maximum power that can be
extracted from the DCM boost rectifier in order to comply with the IEC standard. The harmonic
injection method and the variable switching frequency control strategies are able to decrease the
magnitude of the fifth harmonic, while increasing the amplitude of the other higher-order
harmonics. In fact, the harmonic injection method and the variable switching frequency control
shift the energy concentrated at the fifth harmonic to other higher-order harmonics.
Fig. 2-13(b) shows the high-frequency spectrum obtained from the input currents generated by
the three control strategies under comparison. As expected, different control strategies produce
different high-frequency spectra. The fixed duty cycle control and the harmonic injection method
with fixed switching frequency have very similar high-frequency spectra. However, in the upper
frequency range the harmonic injection method enables the rectifier to slightly reduce the
frequency harmonic over the entire frequency range. This control strategy is advantageous for
reducing the noise because it spreads out the amplitude spectrum across a range of frequencies.
The conducted noise measurement under standard conditions actually considers a window of
frequencies. In this way, the variable frequency control at high frequencies may not exhibit
results that are much better than those of the other two fixed frequency control schemes. Fig.
2-13(c) shows the high-frequency spectra when the quadratic sum of the harmonics within a
9kHz window of frequencies is taken into account. Under this assumption, the gain of the
harmonics that are inside the sweeping window is unity, while the gain of the harmonics outside
the window of frequencies is zero. As a result, when this measurement method is taken into
account, the advantage of the variable frequency control becomes less pronounced in the high
frequency range.
The interleaving of DCM boost rectifiers was demonstrated in section 2.5 as an effective
is highly dependent on the phase-shift between the two interleaved rectifiers. Fig. 2-14 illustrates
the effect of the phase-shift error on the amplitude of the harmonics concentrated around the
switching frequency when two rectifiers are interleaved. If the phase-shift differs ±10% from
180o, the amplitude of the harmonics around the switching frequency already becomes
(2fs±fr). As a result, the interleaving of two DCM rectifiers will not be able to eliminate the input
from 180o. This effect can be observed in the experimental spectrum of the interleaved current
illustrated in Fig. 2-10(d), which shows that despite the interleaving of the two rectifiers, there is
1
10 Harm. Injection
Fixed D, Fs
Variable Fs
IEC61000-3-2
0
10
Magnitude (A)
-1
10
-2
10
2 4 6 8 10 12 14
Harmonic number
(a) (b)
(c)
Fig. 2-13. (a) Low-order harmonic comparison, (b) high-frequency spectrum for each control
strategy and (c) equivalent high-frequency spectrum when a 9kHz window is considered in the
spectrum calculation (quadratic sum).
0
10
Noise at fs (A)
Amplitude amplitude
Maximum of the
dominant harmonic
of second peak at 2fs
-1
10
-2
10
-100 -80 -60 -40 -20 0 20 40 60 80 100
Phase-shift error (%)
Fig. 2-14. The impact of phase-shift error on the amplitude of the harmonic at the switching
frequency for two interleaved rectifiers operated with fixed frequency and fixed duty cycle control.
This section provides the DM input filter design and comparison for the following cases: (1)
CCM single-phase boost rectifier, (2) VIENNA rectifier and (3) interleaved system. The
comparison of DM input filter parameters is provided for two different standards: the VDE 0871
and EN 55022 (CISPR 22), both Class B, as shown in Fig. 2-15(a). The equivalent DM input
filter per phase network used in the comparison is shown in Fig. 2-15(b). The filter is a fourth-
order Cauer-Chebyshev (CC) network in which Rd is the high-frequency damping resistor and Ld
is the low-frequency current bypass inductor used to reduce loss in the damping resistor [56]
[57].
The system parameters used in the comparison are summarized in Table 2-5. The total power
under consideration and throughout the remainder of this dissertation is 6kW. The same boost
inductance was used in both benchmark circuits (CCM boost and VIENNA rectifiers), and was
determined to limit the maximum instantaneous peak-to-peak current ripple to below 25% of the
fundamental input peak current measured at low-line input voltage and full load. Additionally,
80
54 dBµV 48 dBµV
60 50 dBµV
No ise Limi t (dBµV)
40
46 dBµV
20
VDE 08 71 Class B
CISPR 2 2 Class B
0
0.01 0.1 1 10 100
Frequency (MHz)
(a)
Rd L3
Ld L1
L2
C4
C2
(b)
Fig. 2-15. (a) EMI standards for conducted noise (average limits) and (b) equivalent DM filter for
one phase.
Vin (V)
L (µH)
System Power (kW) Vbus (V) Line-to-
@ 40kHz
Neutral
Single-Phase 2 (Need three
CCM Boost rectifiers to 600 400 170
Rectifier supply 6kW)
VIENNA
6 600 800 170
Rectifier
Interleaved
140 (Per
Single-Switch 6 800 170
inductance)
DCM Rectifiers
measurement setup used for single-phase equipment, as shown in Fig. 2-16. The dotted box
represents the line impedance stabilization network (LISN), which is used to provide a well-
defined impedance to the noise source generated by the equipment under test (EUT). As can be
The setup to evaluate the DM EMI noise of three-phase equipment is shown in Fig. 2-17. The
result of the simulation in one of the three phases is used to design the DM filter parameters.
LISN
50µH
1µF 0.1µF
Equipment under
5Ω 1kΩ 50Ω
test (EUT)
LINE Dummy
load
NEUTRAL
Port
RF
5Ω
1kΩ
1µF 0.1µF
50µH
50Ω
Spectrum
Analyzer
1µF 0.1µF
Spectrum
Analyzer
OHM
Equipment under
5Ω 1kΩ
50
test (EUT)
B
LISN
C
LISN
Fig. 2-17. Conducted EMI noise simulation setup for three-phase equipment.
Identifying the worst-case input current ripple is rather important to assure that the DM input
filter parameters are designed to attenuate the maximum noise. Fig. 2-18 illustrates how the most
significant harmonics of the input current ripple change with the output power for both CCM
boost (Fig. 2-18(a)) and VIENNA (Fig. 2-18(b)) rectifiers. As can be verified, both CCM
rectifiers generate similar harmonics for similar converter parameters. These harmonics were
obtained for a switching frequency of 40kHz, but the spectrum shape remains the same when
Fig. 2-19 illustrates the harmonics of the input current ripple for the following cases: (a) a single-
unit non-interleaved DCM single-switch boost rectifier supplying 6kW and (b) a two-channel
6kW interleaved system. As can be seen, the non-interleaved DCM boost rectifier generates a
harmonics around fs and other odd harmonics. Since both systems in Fig. 2-19 process the same
power and use equivalent parameters, the even harmonics are identical in both cases.
1
Harmonics of switching frequency (A)
0.9
0.8
0.7
fs 2fs 3fs 4fs
0.6
Vin=170V
0.5
Vo=400V
0.4 Pmax=2kW
0.3 Fs=40kHz
L=600µH
0.2
0.1
0
0 500 1000 1500 2000
Power (W)
(a)
1.2
Harmonics of switching frequency (A)
0.8
fs 2fs 3fs 4fs Vin=170V
0.6 Vo=800V
Pmax=6kW
0.4 Fs=40kHz
L=600µH
0.2
0
0 1000 2000 3000 4000 5000 6000
Power (W)
(b)
Fig. 2-18. Input current ripple harmonics: (a) single-phase CCM boost rectifier (2kW) and (b)
VIENNA rectifier (6kW total power – 2kW per phase).
1.8 Vin=170V
1.6 Vo=800V
Pmax=6kW
1.4
Fs=40kHz
1.2 L=140µH
1
2xfs 4xfs 6xfs
0.8
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4 0.5
Duty cycle
(b)
Fig. 2-19. Input current ripple harmonics: (a) one non-interleaved DCM boost rectifier (6kW), and
(b) two-interleaved rectifiers (6kW).
Fig. 2-15(a) shows the maximum noise profiles allowed by the VDE 0871 and CISPR 22 Class B
standards (see page 56). The VDE is a difficult standard to comply with because it starts to limit
power electronics equipment must be considered when calculating the filter parameters. In the
calculation of the DM input filter, it is assumed that the common-mode noise generated by the
systems under consideration is neglected. This assumption allows the EMI standard profiles to be
used to limit the emissions of DM noise [55]. Furthermore, the calculations hereafter consider
Table 2-6 illustrates for fs=40kHz the operating point at which the DM filter must be designed to
attenuate the worst-case input current ripple. These operating points have been determined with
the help of the curves illustrated in Fig. 2-18 and Fig. 2-19. For instance, consider the noise
generated by the interleaved system, as shown in Fig. 2-19(b). If the DM input filter were
designed to satisfy the VDE standard at a switching frequency of 40kHz, then one can conclude
that the worst-case ripple would occur at 2xfs, since this frequency would be the first harmonic to
be limited by VDE. More specifically, the worst-case ripple would occur at 2xfs when D=0.25.
Similarly, if the filter were designed according to CISPR, then the worst-case ripple would occur
The comparison of the DM input filter size is provided for switching frequencies that vary from
frequencies each of the cases under comparison is most effective in reducing the size of the input
filter. Besides considering the size of the input filter, it is also important to take into account the
Fig. 2-20(a) shows the boost inductance required by each case as a function of the switching
frequency. For the interleaved system, the curve shown in Fig. 2-20(a) represents each
be related to inductor size, the remaining information shown in Fig. 2-20 illustrates the core
weight, winding and total weight for the boost inductors in each case under comparison.
been used in the design [58], while the temperature rise has been always limited to less than
55oC. Fig. 2-20(b) shows the core material weight needed to implement the boost inductors per
phase (for the interleaved system, the weight shown refers to both boost inductors used per
phase). Despite the lower inductance calculated for the DCM boost rectifiers, the amount of core
is comparable to the CCM boost and VIENNA rectifiers because the core loss increases at high
frequency in the DCM case. Fig. 2-20(c) shows the calculated amount of Cu needed to wind the
inductors. The inductors for the CCM circuits require more Cu because the number of turns is
3
Although the highest noise occurs at Po≈300W (see Fig. 2-18(a)), the design of the DM input filter according to the
VDE standard is done at full load for the CCM boost rectifier. The difference between the noises generated at
2kWand 300W is minimal and should not affect the comparison.
4
The same observation is valid for the VIENNA rectifier. Even though the highest noise occurs at Po≈900W (see
Fig. 2-18(b)), the design of the DM input filter according to the VDE standard is done at full load for the VIENNA
rectifier.
shown in Fig. 2-20(d), which illustrates that the total weight of the inductors used in the DCM
rectifiers is lower than the weight of the inductors used in the CCM rectifiers.
200
DCM boost non-inter
DCM boost non-inter
DCM boost - inter
DCM boost - inter
600 CCM boost
150 CCM boost
VIENNA
VIENNA
400
100
200
50
0
0 50 100 150 200 0
0 50 100 150 200
Switching frequency [kHz]
Switching frequency [kHz]
(a)
(b)
Total boost inductor weight per phase [g]
Boost inductor Cu weight per phase [g]
250 500
DCM boost non-inter DCM boost non-inter
DCM boost - inter DCM boost - inter
200 400
CCM boost CCM boost
VIENNA VIENNA
150 300
100 200
50 100
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(c) (d)
Fig. 2-20. Boost inductor design results: (a) boost inductance, (b) core weight, (c) Cu weight, and
(d) combined core + Cu weight per phase.
The determination of the filter parameters follows the steps described in Appendix II, while the
total capacitance per phase for all cases under comparison can be calculated according to [56]:
I pk −min
(2-32) C max = tan (a cos(DF ) ) ,
2π f r V pk −high
where Ipk-min is the peak line current at light load (usually defined as 15% of the full power), Vpk-
high is the peak line-to-neutral input voltage at high line, and DF is the displacement factor at light
The equivalent filter per phase used in the calculations is shown in Fig. 2-15(b) (see page 56).
The results are presented in Fig. 2-21, which shows the inductances L1 and L3, the filter core
weight and the combined core plus Cu weight to implement the filter inductors. The inductance
L2 is not considered in the calculation because its size does not affect the overall filter, since L2
is not in the main path of the input current. The low-frequency bypass inductor Ld is considered
to be identical to L1, not only in terms of inductance value but also size, while the total
capacitance determined from (2-32) is 3.8µF for all cases. The Kool Mµ material is also used to
calculate the weight of the filter inductors used in the converters under comparison [58].
The impact of the interleaving technique in reducing the filter size under the VDE standard is
clear from Fig. 2-21(c). The non-interleaved DCM boost rectifier requires a huge amount of core
material to implement the filter inductors in order to provide appropriate attenuation. On the
other hand, the interleaved system requires even less filtering than the CCM boost and VIENNA
rectifiers.
2000
CCM boost CCM boost
800 VIENNA VIENNA
1500
600
1000
400
200 500
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
1200 3000
600 1500
400 1000
200 500
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(c) (d)
Fig. 2-21. Filter size for VDE 0871 Class B: (a) filter inductance L1, (b) filter inductance L3, (c) filter
core weight to implement Ld+L1+L3 and (d) Core + Cu weight per phase required to implement
Ld+L1+L3.
Although the interleaving technique was effective in reducing the size of the DM input filter for
the VDE standard, it remains to be determined whether or not the same reduction occurs when
the CIPSR 22 is taken into account. To answer this inquiry, a similar evaluation was carried out
to calculate the weight of the DM input filter using the CISPR 22 Class B standard.
inductance per phase is drastically reduced in all cases as compared to the inductances calculated
for the VDE standard. The second observation is that interleaving does not necessarily reduce the
amount of filter inductance. As a matter of fact, for switching frequencies below 50kHz, the
interleaved system is not even better than the non-interleaved DCM boost rectifier. In addition,
below 50kHz the CCM boost and VIENNA rectifiers require a minimal amount of filter
inductance per phase. The interleaving becomes more advantageous above a switching frequency
Fig. 2-22(c) shows the amount of magnetic core material (Kool Mµ powder core) needed to
implement the filter inductors Ld, L1 and L3 for one phase, while Fig. 2-22(d) includes the Cu
weight in the evaluation. As can be seen, the interleaved system is most advantageous at
frequencies above 150kHz. The CCM and VIENNA rectifiers require smaller filters below
150kHz than any other DCM boost rectifier, with the best points for reducing the DM filter size
defined at 50kHz, 70kHz and 150kHz. The interleaved system presents some advantage over the
non-interleaved DCM boost rectifier for switching frequencies between 50kHz and 75kHz.
Although interleaving does not result in a reduced filtering requirement for the CISPR 22
standard, it is important to remember that the DCM rectifiers require smaller boost inductors
than their CCM counterparts, as shown in Fig. 2-20(d). Therefore, combining the weight of filter
and boost inductors per phase certainly changes the comparison results, as shown in Fig. 2-23.
The combined filter with boost inductor magnetic core weights is shown in Fig. 2-23(a), while
the incorporation of Cu into the evaluation is shown in Fig. 2-23(b). As can be seen, between
boost and VIENNA rectifiers. Above a switching frequency of 150kHz, the interleaved system is
deemed the most effective in reducing the weight of boost and filter inductors.
200 400
DCM boost non-inter DCM boost non-inter
DCM boost - inter DCM boost - inter
Filter inductance L1 [µH]
100 200
50 100
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
300
Core + Cu filter weight per phase [g]
1000
DCM boost non-inter
DCM boost non-inter
250 DCM boost - inter DCM boost - inter
CCM boost 800
Filter core weight [g]
CCM boost
VIENNA
200 VIENNA
600
150
400
100
50 200
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(c) (d)
Fig. 2-22. Filter size for CISPR 22 Class B: (a) filter inductance L1, (b) filter inductance L3, (c) filter
core weight to implement Ld+L1+L3, and (d) Core + Cu weight per phase required to implement
Ld+L1+L3.
followed by a DC/DC converter used to regulate the DC output voltage. In such a configuration,
intermediate bus capacitance must be large enough to decouple both the operation and the
dynamics of the two power conversion stages. For this reason, each stage can be designed and
optimized separately.
300
Filter + boost inductors
200 600
150
400
100
50 200
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
Fig. 2-23. (a) Combined filter and boost inductor core weight per phase and (b) total core + Cu
weight needed to implement the boost and filter inductors per phase.
In single-phase applications, the PFC circuit controls the intermediate bus voltage so that it stays
at 400V, making the full-bridge converter shown in Fig. 2-24(a) the preferred topology for
DC/DC power conversion. Since the bus voltage is regulated at 400V, the use of 500V
MOSFETs in the DC/DC converter is feasible. Devices rated at 500V present very good Rds-on
characteristics, boosting up the efficiency of the entire system. The full-bridge converter is
widely used in power electronics systems for applications above 1kW, and has an extensive list
(a)
+
-
(b)
+
-
(c)
Fig. 2-24. DC/DC power conversion topologies: (a) full-bridge, (b) dual-bridge and (c) three-level.
380V, the intermediate bus voltage is usually controlled to 800V by the PFC stage. Therefore,
the DC/DC converter must handle the intermediate bus voltage stress, while still operating with
DC/DC converter is not so clear in this type of three-phase application because the device
voltage rating must be at least 1000V. For this voltage rating, power MOSFETs would not offer
good Rds-on characteristics, since they would increase the conduction loss in the circuit. On the
other hand, if IGBTs were used in the DC/DC converter, the switching frequency would have to
be lower to avoid high turn-off loss. Reducing the switching frequency would decrease the
power density of the entire system, which is not desirable. Therefore, it is necessary to
investigate different solutions for this type of high-voltage application. In fact, other topologies,
such as the dual-bridge and the three-level DC/DC converters, are potential candidates [13]-[15].
The main advantage of these converters is that 500V devices can still be used to implement the
circuit despite the bus voltage being 800V, as illustrated in Fig. 2-24(b) and Fig. 2-24(c).
To show the major differences between the efficiencies of the full-bridge and the three-level
topologies, Fig. 2-25 shows a theoretical comparison between a three-level converter using 600V
the state-of-the-art power MOSFETs when the comparison was carried out in 2000. Conduction
and switching losses have been taken into account in the power switches, while only conduction
loss was considered in the output rectifiers. As can be seen, the three-level converter using 600V
devices results in better efficiency. It is also important to mention that 600V devices present
lower cost than their 1000V counterparts. Better efficiency and lower device cost justified the
choice of the three-level converter for applications in which the intermediate bus voltage is
800V.
96
Efficiency (%)
94
92
Three-level
90
Full-bridge
88
0 20 40 60 80 100
Output current (A)
Fig. 2-25. Theoretical efficiency calculated for full-bridge and three-level converters.
Soft-switching techniques are necessary in order to improve the converter efficiency at higher
switching frequencies. Purely ZVS converters rely on the energy stored in the resonant
inductance placed in the primary side of the converter to achieve zero-voltage turn-on. A large
resonant inductance increases the load range in which the power switches operate with zero-
voltage turn-on, while increasing the circulating energy in the circuit [62]. In addition, the
resonant inductance placed in the primary side of the DC/DC converter oscillates with the
junction capacitance of the secondary-side rectifier, causing overshoot and requiring snubbers to
damp the oscillations. These problems can be overcome when zero-voltage and zero-current
switching (ZVZCS) techniques are used instead of the purely ZVS schemes. The ZVZCS
techniques also reduce the circulating energy in the primary side of the transformer, providing
zero-voltage turn-on for the outer switches of the three-level topology and zero-current turn-off
for the inner switches, while drastically reducing the secondary-side parasitic ringing across the
output rectifier diodes [63] [64]. To achieve ZVZCS operation, each switch in the three-level
DC/DC converter operates with nearly 50% duty cycle. As shown in Fig. 2-26(a), it is also
guarantee ZVS for the outer switches, as well as connect a secondary lossless circuit to reset the
primary current during the freewheeling period in order to provide ZCS for the inner switches.
Fig. 2-26(b) shows the primary voltage across the transformer (upper trace), the voltage across
the output rectifier (middle trace) and the primary transformer current (lower trace). It can be
observed that the primary current is reset to zero before the inner switches of the DC/DC
converter are turned off, which characterizes a zero-current turn-off for the inner switches. The
leakage inductance resonates with the capacitance of the lossless circuit placed in the secondary
side, which is shown by the resonant peak seen in the primary current waveform. As can be seen,
the voltage across the output rectifier of the three-level DC/DC converter does not present
parasitic ringing. The parasitic ringing was eliminated in the waveform because the ZVZCS
three-level converter does not need the resonant inductance to achieve soft switching for the
inner switches. Therefore, the resonant inductance is simply the leakage inductance of the
transformer, which is the reason for reducing the parasitic ringing across the secondary rectifiers.
Fig. 2-26(c) shows the drain-to-source voltage across the upper switch of the three-level DC/DC
converter and its gate signal. The voltage across the switch decreases to zero before the gate
signal is applied, which characterizes a zero-voltage turn-on for the outer switches of the three-
level DC/DC converter. As can be verified, the switches in the three-level DC/DC converter
support half of the total bus voltage. This allows the use of 600V MOSFETs with low Rds-on in
+ a
b
- Llk
vs
(a)
Vab Vds
Vgs
Vs
ZVS
ILlk
(b) (c)
Fig. 2-26. (a) Three-level ZVZCS DC/DC converter used in the implementation of the two-stage
front-end converter, (b) experimental results for vab, vs and iLlk (20A/div) at Po=5kW, Vbus=800V,
Vout=52V and fs=100kHz, and (c) ZVS transition for the outer switches.
Fig. 2-27 shows two possible methods for realizing a front-end converter using interleaved DCM
boost rectifiers and three-level DC/DC converters. In the first option, shown in Fig. 2-27(a), the
DCM boost rectifiers are connected to the same bus voltage, while one three-level DC/DC
converter is used to process the total output power. The second option, shown in Fig. 2-27(b),
uses two individual channels to process the power. The main difference between the two
output side of the three-level DC/DC converters, while in the first case no current sharing is
needed because the boost rectifiers are connected together and they operate as current sources
(see results in Fig. 2-12(a)). Evidently, the tolerance in the value of the boost inductances of the
first option will dictate the level of current sharing achieved by connecting the DCM boost
rectifiers to the same bus voltage. However, well-defined manufacturing procedures should
guarantee that the inductance values are within a 10% tolerance, which should be sufficient to
guarantee current sharing between the two DCM boost rectifiers connected to the same bus
voltage.
Although the system shown in Fig. 2-27(a) is simpler and does not present current sharing issues,
its counterpart shown in Fig. 2-27(b), using two three-level DC/DC converters, presents a higher
level of modularity. Nevertheless, the DPS front-end converter was implemented using the first
approach, shown in Fig. 2-27(a), whose main components are described in Table 2-7 for the 6kW
application. For this particular example, the PFC stage was implemented at 40kHz of switching
frequency using a ZCT circuit to provide zero-current turn-off for the main IGBT, while the
three-level DC/DC converter was operated at 100kHz. Fig. 2-28 shows the individual converter
and the overall system efficiencies at full load as a function of the line-to-neutral input voltage
variation. The efficiency of the PFC stage changes linearly with the variation of the line-to-
neutral voltage, while the efficiency of the three-level DC/DC converter remains constant over
the entire input voltage variation. This is in agreement with the fact that the intermediate bus
voltage is regulated by the PFC stage. The maximum overall efficiency is achieved at high-line
input voltage, while the overall efficiency at 220V line-to-neutral input voltage is 90.8%.
- vds +
L1 ZVZCS converter
~ EMI 100kHz
L2
~ Input
L3
~ Filter
- vs +
a
iLK
b
L4
L5
L6
(a)
Input Filter
(b)
Fig. 2-27. Front-end converters using interleaved DCM boost rectifiers: (a) common intermediate
bus voltage, and (b) using two DC/DC converters.
98
Single-switch boost
96
DC/DC Converter
Efficiency (%)
94
92
Overall efficiency
90
88
160 180 200 220 240 260 280
Line-to-neutral input voltage (V)
2.9. Benchmarking
As mentioned in chapter 1, the circuits described in this work are benchmarked against the CCM
and VIENNA rectifier. Table 2-8 illustrates the comparison between the benchmark circuits and
the single-switch DCM boost rectifier. The data marked up in blue represent advantages with
respect to benchmark circuits, while read fonts represent disadvantages. From Table 2-8, one
efficiency.
Table 2-8. Benchmarking the single-switch DCM boost rectifier (Vin=220V LN).
DC/DC
Feature
L6
EMI Converter
C Filter
D14 D15 D16
(Full-bridge D20
ZVS)
Total power
6 6 6
(kW)
Switches 3 3 2
Line freq.
diodes 12 12 -
Fast diodes 3 6 16
Bus voltage
400 800 800
(V)
Switch voltage
400 400 800
(V)
RMS/AVG
5.5/3.2 5.5/3.2 4.9/2.3
SW current (A)
AVG diode
current
Line freq.
4.1 4.1 -
Diodes (A)
5 2.5 4.1 (input rec.) – 3.75 (output)
Fast diodes (A)
Output cap
RMS current 5.5 5.6 4.75
(A)
THD (%) - - 12.7
THD with
harmonic - - 10.8
injection
Efficiency (%)
98 (simulation) 98.2 (simulation) 95.8
at 40kHz
Power under
two-phase 4 4 3.4
operation (kW)
Active current
Yes Yes No
control
Sensing effort Medium High Low
Control
High High Low
complexity
This chapter explored and described the pros and cons of using the single-switch DCM boost
rectifier as a PFC stage for applications in DPSs. Detailed analyses and simplified design
guidelines were presented. The harmonic injection method demonstrated improvements in the
quality of the currents drained from the power source. As a benefit of the harmonic injection
technique, it is possible to extract more power from the DCM boost rectifier. For a bus voltage of
800V and an input phase voltage of 220V, the harmonic injection enables the extraction of 8kW
from the single-switch DCM boost rectifier, as opposed to 6kW when no harmonic injection is
As demonstrated throughout this chapter, interleaving DCM boost rectifiers provides high-
frequency input current ripple cancellation. As a result of reducing the amplitude of the input
current ripple, a reduction in filter size was demonstrated when the VDE 0871 standard was used
to design the DM input filter. A comparison between the interleaved system, the non-interleaved
DCM boost rectifier, and two benchmark circuits (CCM boost and VIENNA rectifiers) showed
that interleaving is a very effective method for reducing the size of the DM input filter. The
reduction in filter size was achieved because the VDE standard starts at 10kHz, requiring the
On the other hand, when compared with the benchmark circuits, a reduction in filter size was not
achieved. In fact, in terms of the weight of the magnetic cores needed to implement the filter
inductors, both the non-interleaved DCM boost rectifier and the interleaved system yielded
similar sizes below 50kHz of switching frequency. The interleaved system becomes more
advantageous above 150kHz, but below this frequency the CCM boost and VIENNA rectifiers
interleaved system to show some advantages under CIPSR 22, the size of the boost inductors
must be included in the comparison. In that case, for switching frequency of 50kHz to 75kHz,
the combined weight of the boost and filter inductors for the interleaved system becomes
comparable to both the CCM boost and the VIENNA rectifiers. Above 150kHz, the interleaved
system is deemed the most effective in reducing the size of the DM input filter.
Current sharing between the interleaved rectifiers can be easily achieved because the DCM boost
rectifier operates as a current source. The accuracy level in the current sharing depends upon the
difference between the inductances connected to the same phase. A tolerance of ±10% in the
boost inductances would be accurate enough to guarantee good current sharing between the
interleaved DCM boost rectifiers. To maintain input current ripple cancellation, it must be
assured that there will be no phase-shift error in the gate signals of the interleaved rectifiers. As
demonstrated, a phase-shift error of ±10% in the gate signals of the two interleaved rectifiers is
sufficient to create a DM noise around the switching frequency, which is comparable to the noise
at 2xfs.
A two-stage front-end converter for 6kW applications, based on the interleaved system
connected in series with a three-level DC/DC converter, was tested. The system achieved an
overall efficiency of 90.8%, which included the losses in the DM input filter. Despite the
simplicity and excellent overall efficiency of the solution discussed in this chapter, there are still
two drawbacks to overcome: the voltage stress across the switch of the PFC circuit and the
3.1. Introduction
Up to this point, three major methods for obtaining high-power front-end converters with PFC
function were discussed: (1) the connection of single-phase modules to the three-phase system in
order to achieve the required output power level [9]-[11], (2) the VIENNA rectifier [18], and (3)
the single-switch DCM three-phase boost topology [42], including interleaved DCM boost
rectifiers [51]. In the last approach, the boost inductors operate in DCM to achieve automatic
input current shaping. The harmonic injection method was used in the single-switch DCM boost
rectifier to increase the output power level while still meeting the IEC 61000-3-2 standard
without increasing the output voltage beyond 800V [47]. Despite the improvements made in the
operation of the single-switch DCM boost rectifier to avoid increasing the bus voltage beyond
practical levels, the use of power MOSFETs with low on-resistance is not yet possible. To
overcome this problem, this chapter presents a two-switch boost rectifier implemented in a three-
level topology used to reduce the voltage stress applied across each switch. As a result of the
reduced voltage, low on-resistance MOSFETs can be used in the power stage.
This chapter explores the main features of the two-switch three-level boost rectifier operated in
DCM. The main contributions are reduced voltage stress, improved efficiency and reduced THD.
Control strategies and design guidelines are provided throughout the text. The interleaving
technique is also used to cancel the high-frequency input current ripple. The discussion is
A similar version of the topology discussed hereafter, using variable-frequency control, has been
previously presented [65]. The two-switch three-phase rectifier is shown in Fig. 3-1. The three-
level structure is comprised of the switches S1 and S2 and diodes D1 and D2. The capacitors C1
and C2 share the total bus voltage. The AC capacitors Ca, Cb, and Cc eliminate the neutral point
connection of the power system, thus preventing any zero-sequence-order harmonic (3rd, 9th, 15th,
etc…) from circulating in the input lines. The advantage of the three-level structure is that the
voltage applied across the power switches is half of the total bus voltage, which enables the use
D1
vin S1 C1 R1
La v1
iin Lb
Lc
S2 C2 R2 v2
vc
Ca Cb Cc
D2
Fig. 3-2 shows the operating stages of the two-switch three-level boost rectifier, while Fig. 3-3
illustrates the boost inductor current waveforms. To simplify the explanation, Fig. 3-2 neglects
the AC capacitors. During one operating cycle, the switches S1 and S2 are simultaneously turned
on to store energy in the input inductors (first stage). When the switches are turned off, the boost
connected across the same phase of the boost inductor (second, third and fourth stages). In the
fifth stage, which is not shown in Fig. 3-2, the output load is supplied by the energy stored in the
bus capacitors.
3.3. Control Strategy and Voltage Balance Across the Bus Capacitors
Fig. 3-1 also shows a possible way to connect the output load to the rectifier. Instead of
connecting a single load across the positive and negative DC rails, the load can be split between
the output capacitors. Although this approach to connecting the output load is possible, if the
converter is not properly controlled, any difference between the DC currents circulating through
the split output loads results in an offset in the voltages across the AC capacitors, as well as in an
imbalance of the voltages across C1 and C2. Fortunately, this voltage imbalance across the bus
One possible scheme for controlling the voltage imbalance across C1 and C2 is presented in Fig.
3-4(a). As can be seen, S1 is used to control the voltage across C2, whereas S2 is used to control
the voltage across C1. Both voltage controllers in Fig. 3-4(a) use the same voltage reference,
which is Vbus/2.
Fig. 3-4(b) shows the simulation results obtained from a circuit that supplies a total power of
6kW at 170V line-to-neutral input voltage. The capacitance C1 was set to 400µF and C2 to
500µF, while the equivalent resistor R1 draws 3.3kW from the rectifier and R2 2.7kW. The
S1 C1 S1 C1
La + Va La + Va
~ ~
Lb + Lb Vb +
Vb Ro Ro
~ ~
Lc + Vc Lc + Vc
~ ~
S2 C2 S2 C2
D4 D4
(a) (b)
D3 D3
S1 C1 S1 C1
La + Va La + Va
~ ~
Lb Vb + Lb Vb +
~ Ro ~ Ro
Lc + Vc Lc + Vc
~ ~
S2 C2 S2 C2
D4 D4
(c) (d)
Fig. 3-2. Operating stages: (a) first stage (to, t1), (b) second stage (t1, t2), (c) third stage (t2, t3), and (d) fourth stage (t3, t4).
ib
S 1 , S2 S1, S2
to t1 t2 t3 t4 t5
As can be verified in Fig. 3-4(b), the voltages across C1 and C2 are controlled to the targeted
value, despite the imbalance in the split load and the mismatch between C1 and C2. It is
important to mention, however, that there is no means of regulating both voltages under severe
load imbalance. For instance, suppose that R2 is suddenly disconnected from the rectifier. In this
case, the control signal that generates the duty cycle for S1 is driven to zero. However, S2 is still
able to control the voltage across C1 to 400V. Therefore, there is still current flowing through
capacitor C2 during the turn-off stage of S2. Since R2 has been disconnected, it is no longer
possible to assure the charge balance in C2, and the converter will not operate properly.
Consequently, under a situation of extreme imbalance, such as the disconnection of one of the
C2
S2 v2
PWM G1(s)
-
+
+ Vbus/2
PWM G2(s)
-
(a)
420
v1
Capacitor voltages (V)
400
v2
380
360
0 10 20 30 40 50
Time (ms)
(b)
Fig. 3-4. Control of voltages across C1 and C2: (a) control scheme and (b) simulation results.
The flexibility of using S1 to control v2 and S2 to control v1 enables the use of the two-switch
three-level DCM boost rectifier, as illustrated in Fig. 3-5(a). Since the voltages across C1 and C2
can be individually controlled, conventional DC/DC converters usually designed to operate from
3-5(b), the two-switch boost rectifier supplies power to a three-level DC/DC converter. Any duty
cycle mismatch in the switches of the three-level DC/DC converter causes a voltage imbalance
across C1 and C2. This voltage imbalance, however, can be mitigated by the modulation scheme
D1
DC/DC
vin S1 C1 Converter
La
iin Lb
Lc
S2 C2 DC/DC
Converter
vc Ca Cb Cc
D2
(a)
D1
vin S1 C1
La
iin Lb
Lc
S2 C2
vc
Ca Cb Cc
D2
(b)
Fig. 3-5. Possible applications: (a) using two split DC/DC converters and (b) using a three-level
DC/DC topology.
As shown in Fig. 3-5(a), if DC/DC converters are connected across the output capacitors of the
three-level DCM boost rectifier, they can be used to balance the voltages across C1 and C2. The
the DC/DC converters is illustrated in Fig. 3-6(a). As shown, a voltage controller is used to
regulate the total bus voltage of the rectifier. The task of balancing the voltages v1 and v2 is then
achieved by the two DC/DC converters. The control objective is to maintain the voltage
difference ∆v=v1-v2 to as close to zero as possible. The output of the voltage balance regulator is
added to or subtracted from the current reference of the DC/DC converters. As a result, both
DC/DC converters will draw the right amount of power to maintain the voltage balance across
the bus capacitors. This situation is shown in Fig. 3-6(b), where the bus capacitors have been set
to C1=400µF and C2=500µF. As shown in Fig. 3-6(b), after enabling the voltage balance
compensator at t=20ms, both voltages across C1 and C2 will be regulated to 400V despite the
mismatch in the bus capacitance values. The simulation results show that the voltage balance
method discussed in this paragraph is rather effective in balancing the voltages across the bus
capacitors.
Deriving design guidelines from the simplified circuit shown in Fig. 3-2 is relatively simple.
However, in the practical circuit, the AC capacitors used to provide the artificial neutral point
connection modify the voltage gain of the three-level boost rectifier, thus making it difficult to
obtain a closed-form solution that enables one to obtain the design curves, as demonstrated in
chapter 2 for the single-switch DCM boost rectifier. To simplify the problem of obtaining the
voltage gain when the AC capacitors are connected to the circuit, a time-domain simulation can
DC/DC
C2 v2 Converter
#2
v1+v2 −
− − +
Vbus + PWM H2(s) Gv(s)
Gd(s) PWM +V
− ref
−
+
PWM H1(s)
PFC Converter Control +
∆v=v1-v2 +
Gb(s)
−
∆v=0
(a)
410
Bus cap voltages (V)
405
v2
400
395
390
v1
385
380
0 10 20 30 40 50
Time (ms)
(b)
Fig. 3-6. Using the DC/DC converters to control the voltage imbalance across the DC capacitors of
the three-level DCM boost rectifier: (a) circuit diagram and (b) voltages across C1 and C2.
The AC capacitors must be incorporated by the DM input filter. Therefore, typical values for the
AC capacitances will range from 1µF to 3µF. Fig. 3-7 shows the voltage gain as a function of the
normalized output current obtained by simulation when the rectifier is operated at heavy load.
adjusted to provide critical conduction at full load (6kW) and high-line input voltage (265V
RMS line-to-neutral input voltage). Considering this approach, the rectifier operates as close as
possible to the boundary line between CCM and DCM, which helps to reduce current stress in
the devices. After adjusting the input inductance value, the input voltage was swept from high to
low line (170V RMS line-to-neutral input voltage), while the output bus voltage was fixed at
800V. The duty cycle was then adjusted to provide full output power at a given input voltage.
The results were then normalized and plotted in Fig. 3-7. Although the results shown in Fig. 3-7
were obtained for a specific power case, they can be directly used to design the two-switch three-
level rectifier for any specification because the results have been normalized. The voltage gain
Vo I L fs
(3-1) M= and I norm = ,
V pk V pk
where Vo is the output bus voltage, Vpk is the peak line-to-neutral input voltage, I is the average
The design curve shown in Fig. 3-7 can be used to calculate the input inductance of the three-
level DCM boost rectifier. As an example, suppose that the rectifier is designed to supply
3.15kW at 40kHz of switching frequency. The bus voltage is 800V, and the line-to-neutral input
voltage is allowed to change from 170V to 265V RMS. From the specifications, the voltage gain
of the PFC circuit at high-line input voltage is 2.14. From Fig. 3-7, the normalized output current
results in 0.056, and from (3-1) the value of the boost inductance results in 133µH.
Voltage gain
2.5
2
0.05 0.06 0.07 0.08 0.09
Normalized output current
Fig. 3-7. Voltage gain versus normalized output current at heavy load.
A prototype has been implemented and tested in order to validate the previously discussion. The
two-switch three-level DCM boost rectifier shown in Fig. 3-1 has been implemented with the
following devices: DSEI 30-10A (1000V/30A) for the three-phase input bridge diodes, DSEI 30-
10A (1000V/30A) for the boost diodes, and APT60M75JVR (600V/62A) for the power
MOSFETs. Although the total bus voltage is 800V, the three-level structure of the DCM boost
rectifier enables the use of 600V devices or even 500V devices if the voltage margin is
considered to be sufficient.
The current waveform in one of the boost inductors at 3kW of output power and 180V of input
phase voltage is shown in Fig. 3-8(a). The DCM operation can be observed from the boost
inductor current waveform. The three-level structure of the DCM boost rectifier is able to reduce
(a) (b)
Fig. 3-8. Experimental results: (a) boost inductor current (10A/div) at 3kW and Vin=180V and (b)
voltage across one of the switches.
Fig. 3-9(a) shows the efficiency for two different designs: one at 40kHz and another at 70kHz.
Since MOSFETs are used, the converter should be able to operate at higher frequencies as
compared to the previous chapter in which IGBTs were switched at 40kHz. Including the input
filter in the efficiency measurements, the efficiency is reasonably high for both switching
frequencies. At nominal input phase voltage (220V line-to-neutral voltage), the efficiency of the
two-switch three-level boost rectifier drops only 1% at 70kHz, as compared to the efficiency at
40kHz. The same figure also compares the efficiency of the two-switch three-level boost rectifier
against the efficiency of the single-switch DCM boost topology discussed in the previous
chapter. As can be seen, the two-switch three-level DCM boost rectifier outperforms its
counterpart circuit.
Fig. 3-9(b) shows the THD of the input currents as a function of the input phase voltage. The
same figure also compares the THD of the two-switch three-level boost rectifier against the THD
three-level boost rectifier also presents lower THD. At nominal input phase voltage (220V), the
All the results illustrated in Fig. 3-8 and Fig. 3-9 show that the two-switch three-level DCM
boost rectifier presents superior performance as compared to its counterpart discussed in the
previous chapter.
In order to decrease the THD to less than 10% over the entire input voltage variation (170V –
265V), the bus voltage of the two-switch three-level DCM boost rectifier must be increased to
890V. At this bus voltage, the 600V power switches can still be used with a considerable voltage
margin. However, it is clear that increasing the bus voltage to reduce the THD of the input
currents imposes serious restrictions on the bus capacitors, since they are usually chosen to
support at most 450V. To overcome this problem, the harmonic injection technique can be
applied to the two-switch three-level DCM boost rectifier to reduce the harmonic distortion of
the input current [47]. A diagram of the system implementation using the harmonic injection
method is shown in Fig. 3-10(a). According to what was presented previously [47], if the
sinusoidal voltage across phase a is taken as the reference voltage for the three-phase system,
then the modulated duty cycle applied to the two-switch boost rectifier can be written as
98
Two-switch - 40kHz
Efficiency (%)
97
96
Two-switch - 70kHz
95
Single-switch - 40kHz
94
93
160 180 200 220 240 260
Input line-to-neutral voltage (V)
(a)
25
20
THD (%)
15
Single-Switch
10
Two-Switch
5
0
160 180 200 220 240 260
Line-to-neutral input voltage (V)
(b)
Vd 3π
(3-2) dc =
Vramp 1 + mh sin 6 ω t + 2 ,
where mh is the modulation index of the harmonic injection signal, Vd is the output of the voltage
compensator, and Vramp is the peak voltage of the saw-tooth waveform generated by the PWM
(I 5 − mh I1 )2 − (mh I1 )2
(3-3) THD = ,
I1
where I1 is the fundamental input current and I5 is fifth harmonic measured before the
introduction of the harmonic injection. The minimum THD of the input currents can be achieved
when
1 I5
(3-4) mh = .
2 I1
The main objective of using the harmonic injection method is to reduce the THD at high line. For
the two-switch three-level DCM boost rectifier, the RMS fundamental current at 3.15kW and
265V is I1=4.15A, while the fifth harmonic measured before the application of the harmonic
injection method was 0.62A. From (3-4), mh is set to 7.5% to minimize the THD at high line.
Once mh has been set to provide the minimum THD at high-line input voltage, the harmonic
injection circuitry that is shown in Fig. 3-10(a) will then self-adjust the modulation index mh for
different operating points. The benefit of using the harmonic injection method in the two-switch
three-level DCM boost rectifier can be seen in Fig. 3-10(b). As illustrated, the THD is reduced to
10% at high line, which represents a reduction of approximately five points with respect to the
case without the harmonic injection implementation. At low-line input voltage, the THD of the
two-switch three-level DCM boost rectifier using the harmonic injection method increases
slightly, since the system becomes overcompensated by the self-adjusting modulation index.
injection method is applied, the THD remains below 10% at high line, while still maintaining the
Three-Level
Two-Switch
Rectifier
dc
PWM
Sixth-Order vc -
Vd
Harmonic ++ G(s) +
+ Vref
Generator vm +
k
va
x
(a)
25
Single-switch
20 Two-switch
Harmonic Injection
15
THD (%)
10
0
170 180 200 220 240 260
Line-to-neutral input voltage (V)
(b)
Fig. 3-10. The two-switch three-level DCM boost rectifier using the harmonic injection method: (a)
implementation and (b) THD.
The converter discussed up to this point operates in DCM, and for this reason requires a large
input filter to attenuate the input current ripple. Similar to what was discussed in the previous
chapter, the interleaving technique can also be applied to reduce the amplitude of the input
current ripple. An interleaved configuration using two rectifiers is shown in Fig. 3-11(a).
Assuming that the interleaved system supplies 6.3kW of total power, according to section 3.4
Fig. 3-11(b) shows the amplitude of the main high frequency harmonics for a non-interleaved
6kW three-level DCM boost rectifier, while Fig. 3-11(c) shows the amplitude of the high-
frequency harmonics when interleaving is applied to both three-level DCM rectifiers. The odd
harmonics of the switching frequency are cancelled out by the interleaving effect. Therefore, the
input filter can be designed to have a higher cutoff frequency in order to attenuate the current
ripple. As can be observed, the effectiveness of the input current ripple cancellation is strongly
Fig. 3-12 shows the interleaved current waveform (iL11+iL12) at heavy load and for three different
input phase voltages. The ripple cancellation provided by the two-channel interleaved three-level
DCM boost rectifiers is quite effective. It can also be seen that there is a reduction in the ripple
cancellation effectiveness as the input phase voltage increases, since the duty cycle decreases at
L11
S1 C1
EMI L21
R Vo
filter L31
S2 C2
D2
L12 D3
S3
L22
L23
S4
D4
Ca Cb Cc
(a)
10
Harmonics of switching freq. (A)
2.5
9
8 2
fs 2fs 3fs 4fs
7
6 1.5
5
4 1
2xfs 4xfs 6xfs
3
2 0.5
1
0 0
0.05 0.12 0.18 0.25 0.32 0.38 0.45 0.05 0.12 0.18 0.25 0.32 0.38 0.45
Duty cycle Duty cycle
(b) (c)
Fig. 3-11. (a) DCM interleaved two-switch boost rectifiers, (b) high-frequency harmonics without
interleaving, and (c) high-frequency harmonics with interleaved operation.
220V
260V
10A/div
Fig. 3-12. Interleaved current at 6.3kW (iL11+iL12) (10A/div all traces) at three different input phase
voltages.
In this section, the parameters of the DM input filter are calculated and compared with those for
the single-switch topology and both benchmark circuits (CCM boost and VIENNA rectifiers).
Table 3-1 defines the operating point used for the DM filter design. The operating points for the
non-interleaved and interleaved two-switch three-level DCM boost rectifiers are related to the
worst-case input current ripple, and they have been extracted from the curves shown in Fig.
Fig. 3-13 shows the comparison of the DM filter parameters designed to attenuate the DM noise
according to the VDE 0871 standard. The filter network used in the design is the same one
introduced in the previous chapter (see page 56). The same assumptions are also taken into
calculation of the filter parameters, while the total filter capacitance per phase is 3.8µF.
that interleaving makes a big difference in reducing the amount of filter inductance in order to
provide the appropriate attenuation for the DM noise. Comparing the interleaved system based
on the single-switch DCM boost rectifier with the one based on the two-switch three-level DCM
boost rectifier, the results are quite similar. In fact, the interleaved system based on the three-
level DCM boost rectifier requires slightly more filter inductance at low switching frequencies
because the boost inductance of system studied in the previous chapter is 140µH, while the boost
inductance of the two-switch three-level DCM boost rectifier is 133µH. This difference is
reflected in a greater input current ripple for the two-switch three-level DCM boost rectifier,
which explains the slight difference between the two interleaved systems.
Filter inductance L3 [µ H]
2000
Single-switch inter Single-switch inter
1000 VIENNA VIENNA
1500
1000
500
500
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
Single-switch inter
800 VIENNA Single-switch inter
2000 VIENNA
600
1500
400
1000
200
500
0
0 50 100 150 200 0
Switching frequency [kHz] 0 50 100 150 200
Switching frequency [kHz]
(c)
(d)
Fig. 3-13. Filter size for VDE 0871 Class B: (a) filter inductance L1, (b) filter inductance L3, (c) filter
core weight to implement Ld+L1+L3, and (d) Core + Cu weight per phase required to implement
Ld+L1+L3.
The results of the evaluation are illustrated in Fig. 3-14. As can be seen, in terms of filter
inductance and filter core weight, below a switching frequency of 150kHz, the VIENNA rectifier
requires the smallest filter to attenuate the DM noise, while there is no clear benefit of
interleaving the two DCM rectifiers. Between 50kHz and 150kHz of switching frequency, the
beneficial for reducing the filter size above 150kHz. Although above 150kHz the interleaved
systems require the least amount of filter inductors, the result is not as good as the VIENNA
200 400
Two-switch non-inter Two-switch non-inter
Two-switch inter
Two-switch inter
150 Single-switch inter 300 Single-switch inter
VIENNA VIENNA
100 200
50 100
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
300 1000
Core + Cu filter weight per phase [g]
50 200
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(c) (d)
Fig. 3-14. Filter size for CISPR 22 Class B: (a) filter inductance L1, (b) filter inductance L3, (c) filter
core weight to implement Ld+L1+L3, and (d) Core + Cu weight per phase required to implement
Ld+L1+L3.
inductor is combined with the size of the filter inductors, as shown in Fig. 3-15. The core weight
for the boost and filter inductors is shown in Fig. 3-15(a). As can be seen, the interleaved
rectifiers result in the lowest weight above 150kHz. However, for the switching frequency range
between 50kHz and 75kHz, the interleaving technique also shows some benefit in reducing the
combined boost and filter inductors sizes. In fact, designing the interleaving system so that it is
slightly below 75kHz would be the best design point for minimizing the combined weight of the
boost and filter inductors, since above 150kHz the switching frequency increases to such a point
that soft-switching techniques must be implemented in the power switching stages. Fig. 3-15(b)
includes the Cu weight needed to wind the boost and filter inductors. As can be observed, the
conclusions remain the same when the Cu weight is included in the evaluation.
1000
Total core + Cu weight per phase [g]
300
Filter + boost inductors
200 600
150
400
100
50 200
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
Fig. 3-15. (a) Combined filter and boost inductor core weight per phase and (b) total core + Cu
weight needed to implement the boost and filter inductors per phase.
The results discussed up to this point have been based on the DM noise only. However, it is also
important to understand the mechanisms under which the common-mode noise is generated and
The identification of the disturbance sources and the propagation paths are essential steps
towards understanding differential and common-mode noises in power electronics systems [66].
In this section, three different rectifiers are compared in terms of common-mode noise
generation: (1) the single-phase CCM boost rectifier, (2) the VIENNA converter and (3) the two-
switch interleaved DCM system. To compare the common-mode noise, a simulation was
performed for the three circuits mentioned above by including parasitic elements in the
component modeling. The next sub-sections describe the assumptions taken to model the various
The boost inductors were modeled using PEmag, an Ansoft software component developed to
model inductors and transformers [67]. The inductor model provided by PEmag is shown in Fig.
3-16. The equivalent model assumed for an inductor is the parallel connection of the inductance
and the equivalent capacitance between turns. The boost inductor used in the CCM rectifiers
(CCM boost and VIENNA) was designed at 40kHz using Micrometals T400/8 core size/material
type [68], and having 103 turns wound with AWG 13 single strand Cu wire. Similarly, each
boost inductor used in the two-switch three-level interleaved system was based on the T300/8
shown in Fig. 3-16 reflects the actual aspect ratio for each inductor. The Micrometals powder
core was chosen to design the boost inductor because this material and the core sizes are already
part of the PEmag library. Notice that both CCM rectifiers need one inductor per phase, while
the interleaved system needs two inductors per phase to realize the circuit (two interleaved
rectifiers). The PEmag simulation gives the parallel capacitance shown in Fig. 3-16.
31.1pF
59.42pF 130µH
(b)
600µH
(a)
Fig. 3-16. Equivalent model for each boost inductor: (a) single-phase CCM boost and VIENNA
rectifiers and (b) two-switch three-level interleaved system.
The parasitic capacitance that connects the disturbance/noise source to the ground constitutes the
major propagation path for the common-mode noise. The parasitic capacitance to ground
depends heavily upon the circuit layout and how the power devices are mounted in the system.
The assumption taken in this part of the analysis is that the power devices are mounted on the
heat sink, which is connected to the ground, as illustrated in Fig. 3-17. To simplify the
comparison, it is also assumed that all switches and diodes are packaged using the TO-247
standard package. Moreover, the devices are electrically isolated with a layer of mica. Having
said that, the parasitic capacitance between the base plate of the device and the heat sink can be
Semiconductor
Thermal Insulation
C
Heat Sink
εo εr A
(3-5) Cp = ,
d
where εo is the permittivity of the air (8.85415 x 10-12 F x m), εr is the relative permittivity of the
mica (4.5), A is the area of contact, and d is the thickness of the mica layer (0.25mm). From the
The disturbance sources for all cases under comparison are the voltages across the power
switches. The dv/dt of these voltages are responsible for generating common-mode currents that
flow through the parasitic capacitance. Fig. 3-18(a) shows the equivalent circuit of the single-
phase CCM boost rectifier, including the parasitic capacitance of the boost inductor and the
parasitic capacitance from the base plate of the MOSFET to the ground of the system. Every
point oscillating at high frequency with respect to the ground is a potential propagation path for
the common-mode noise current [66]. Taking this observation into account, the common-mode
current will flow through the parasitic capacitance of the power MOSFET to the ground, as
shown in Fig. 3-18(a). The VIENNA rectifier with the main parasitic components is illustrated in
Fig. 3-18(b). As can be seen, the VIENNA rectifier presents many more possible paths for the
common-mode noise, since all the points connected to parasitic capacitances oscillate at high
frequencies during circuit operation. Finally, Fig. 3-18(c) shows the parasitic components
(a)
(b)
LISN
LISN
LISN
(c)
Fig. 3-18. Parasitic components: (a) single-phase CCM boost rectifier, (b) VIENNA rectifier and (c)
two-switch three-level interleaved system.
• All circuits employed the same amount of Si per phase. For instance, the single-
phase CCM boost rectifier used four switches in parallel. Since three single-phase
modules are needed to build the system, the total number of switches is 12. To
match the Si amount used in the CCM single-phase boost rectifier, the VIENNA
rectifier was simulated with four switches in parallel per phase leg and the
interleaved system with three switches in parallel to match a total of 12 switches.
The MOSFET used in the simulation was the IRF540.
• It is assumed that the drain of the power MOSFET and the cathode of the diodes
are connected to the base plate of the TO247 package. This assumption is
important because it defines the connection point of the parasitic capacitance from
each device to the ground.
• Based on the number of switches connected in parallel, the gate driver speed was
adjusted to be the same for all cases. In this way, it is guaranteed that the dv/dt
during fall and rise time is also the same.
In order to compare the levels of common-mode (CM) noise generated by each case under
comparison, the DM and CM noises should be separated from the total noise picked up by the
50Ω LISN impedance. Fig. 3-19(a) shows the equivalent setup used in the simulation of the
single-phase CCM boost rectifier. From that equivalent setup, the DM and CM noises can be
determined by
In a similar fashion, Fig. 3-19(b) shows the equivalent setup used in the simulations of the three-
phase rectifiers. The separation of DM and CM noises in each phase can be obtained using the
following relationships:
1
Va − DM = (2 v1 − v2 − v3 )
3
1
Vb− DM = (2 v2 − v1 − v3 )
(3-7) 3
1
Vc − DM = (2 v3 − v1 − v2 )
3
VCM = −(v1 − v2 − v3 )
Although the phase angle for each DM noise harmonic is different because of the three-phase
system, the amplitude of the DM noise in each input phase is the same.
LISN
LISN IDM Ia-DM
EUT Ib-DM EUT
1φ Ic-DM 3φ
ICM
ICM
v1 v2 v1 v2 v3
(a)
(b)
Fig. 3-19. Equivalent high-frequency noise path: (a) single-phase and (b) three-phase.
generates the lowest level of CM noise, while the interleaved system generates the largest level
at 2xfs. Notice that the interleaved system has a concentrated harmonic at 2xfs, while the CCM
boost and VIENNA rectifiers show a more scattered spectrum of harmonics because of the
severe reverse recovery of the fast diodes, as opposed to the interleaved system that presents no
reverse recovery, explaining why the spectrum is more concentrated at 2xfs. The CCM boost
rectifier presents nearly the same noise amplitude between 80kHz to 400kHz, while for the
VIENNA rectifier this frequency range occurs between 80kHz and 300kHz. Although the
previous DM input filter comparison results showed that it is more advantageous to design the
interleaved system between 50kHz and 75kHz, the same results also presented limited meaning
because the CM filter has not been taken into account in the comparison. However, it is clear
from the simulation results shown in Fig. 3-20 that the interleaved system also requires lower
Fig. 3-21 illustrates three categories of device stress: average switch current, RMS switch
current, and RMS bus capacitor current. As can be seen, the switches of the interleaved systems
present similar average and RMS currents, as compared to the CCM boost and VIENNA
rectifiers. Regarding the voltage stress, the single-switch DCM boost rectifier must support the
highest voltage stress (800V). The CCM boost PFC and the VIENNA rectifier can be realized
with 500V MOSFETs, while the interleaved system based on the two-switch three-level DCM
100
CM (dBµV) CM (dBµV)
90
CCM boost 120
VIENNA
80 100
70
80
60
50 60
40
40
30
(a) (b)
20 20
195
CM (dBµV)
170
Two-Switch Three-Level
145
120
95
70
45
(c)
20
Fig. 3-20. CM noise generated by the various converters: (a) CCM boost, (b) VIENNA and (c) interleaved.
As can be verified, both interleaved systems present the lowest RMS bus capacitor currents of all
6 10
Average switch current (A)
1 2
160 180 200 220 240 260 280 160 180 200 220 240 260 280
Line-to-neutral RMS input voltage (V) Line-to-neutral RMS input voltage (V)
CCM boost and VIENNA rectifiers CCM boost and VIENNA rectifier
Single-switch DCM boost interleaved Single-switch DCM boost interleaved
Two-switch DCM boost interleaved Two-switch
- DCM boost interleaved
(a) (b)
8
Bus capacitor RMS current (A)
3
160 180 200 220 240 260 280
Line-to-neutral RMS input voltage (V)
CCM boost PFC
VIENNA rectifier
Single-switch
- DCM boost interleaved
Two-switch DCM boost interleaved
(c)
Fig. 3-21. Device stress: (a) average switch current, (b) RMS switch current and (c) RMS bus
capacitor current.
Since the 2kW single-phase CCM boost PFC and the 6kW VIENNA rectifier were not
prototyped for operational tests, a viable method for comparing the efficiency was to perform a
simulation of the converters based on the loss models developed for MOSFETs and diodes using
The assumptions made are similar to those in the previous section. For instance, the same
amount of Si was used per phase in all the converters, which resulted in a total of 12 switches
evenly distributed among the phases of each approach. The gate driver was then adjusted to
reproduce the same switching speed in order to make a meaningful comparison of switching
losses. As seen in Fig. 3-22, the interleaved system performs very well as compared to the CCM
and VIENNA rectifiers. The core and winding losses of the boost inductors were not included in
the evaluation results illustrated in Fig. 3-22, while the switching frequency used in the
99
VIENNA
98.5
Two-Switch
98
Efficiency (%)
97.5 Single-Phase
CCM Boost
97
96.5
96
95.5
170 190 210 230 250 270
Line-to-neutral input voltage
The ultimate objective is to build a two-stage system that can supply 6kW of power at 48V of
DC output voltage for telecom applications using the two-switch interleaved system in the PFC
stage. Fig. 3-23(a) shows the front-end converter implemented with the two-switch three-level
interleaved DCM boost rectifiers, followed by a three-level DC/DC converter implemented with
the ZVZCS technique briefly discussed in the previous chapter. Fig. 3-23(b) shows the
experimental efficiency comparison between the system developed in this chapter and the one
developed in the previous chapter using the interleaved single-switch DCM boost rectifier. Even
when operating the two-switch three-level interleaved system at 70kHz, the overall efficiency of
the system (including the DM input filter) is still higher than the efficiency of the two-stage
converter discussed in the previous chapter. Moreover, a switching frequency of 70kHz has been
demonstrated as one of the best switching frequencies for reducing the size of the combined
3.13. Benchmarking
Table 3-2 shows a comparison summary between the benchmark circuits and both DCM boost
rectifiers studied to this point. As can be seen, the two-switch three-level DCM boost rectifier
presents substantial improvements over the single-switch DCM boost rectifier in terms of switch
voltage, efficiency, and THD. The DCM boost rectifiers also present reduced RMS bus capacitor
current, while the power transferred under two-phase operation is almost the same as the
benchmark circuits. The other points in favor of the DCM boost rectifiers are the sensing effort
but it is not yet possible to achieve the performance of the benchmark circuits.
Interleaved rectifiers
Interleaved rectifiers
40kHz
70kHz D1 DC/DC three-level
ZVZCS converter
L1
S1
100kHz
EMI L2 C1
filter L3
S2
D2
D3
L4
S3
L5 C2
L6
S4
D4
Ca Cb Cc
(a)
94
93
Efficiency (%)
92
91
90
Single-switch at 40kHz + DC/DC at 100kHz
89 Two-switch at 40kHz + DC/DC at 100kHz
Two-switch at 70kHz + DC/DC at 100kHz
88
160 180 200 220 240 260 280
Line-to-neutral RMS input voltage (V)
(b)
Fig. 3-23. (a) Two-stage front-end converter and (b) experimental efficiency comparison.
Feature EMI S4
C Converter D14 D15 D16 D4
Filter (Full-bridge D20 Ca Cb Cc
ZVS)
This chapter presented the two-switch three-level DCM boost rectifier as an alternative for PFC
in DPS applications. The advantage of three-level topologies is the 50% reduction of the voltage
applied across the power switches. As a result, lower-voltage devices such as MOSFETs with
low on-resistance can be used to improve the rectifier performance. MOSFETs can also operate
at higher switching frequencies in order to reduce the size of magnetic components, such as the
97.4% at 40kHz and 96.4% at 70kHz. Both efficiencies, measured at 220V of input phase
voltage, outperform the efficiency of its counterpart single-switch DCM boost rectifier, which
achieved 95.8% at 40kHz. The THD of the two-switch three-level DCM boost rectifier is almost
five points lower than the THD of the single-switch DCM boost rectifier. It was shown that the
implementation of the harmonic injection method helped to reduce the THD to below 10% over
The interleaving technique also demonstrated benefits in reducing the DM filter size of the two-
switch three-level DCM boost rectifier. The DM filter size reduction achieved by interleaving
two rectifiers was clear when the filter was designed to attenuate the DM noise according to the
VDE 0871 Class B standard. However, when the CISPR 22 was used, the benefit of filter size
reduction was not as obvious, except when the switching frequency was increased above
150kHz. Below this switching frequency, the VIENNA rectifier was clearly more advantageous.
The combined size of the boost and filter inductors then showed that interleaving the DCM boost
rectifiers is indeed effective in reducing the requirements for magnetic components. In fact,
point at which to design the interleaved system in order to minimize the size requirements for the
boost and filter inductors. Between 75kHz and 150kHz, there are no advantages in using
interleaving to reduce magnetic component size. Above 150kHz, the interleaving of two DCM
boost rectifiers again becomes attractive, but then the higher switching losses that occur due to
4.1. Introduction
The discussions of the previous chapters revolved around the two-stage approach for front-end
converters, which started with single-phase modules to configure a high-power system, and then
progressed towards three-phase approaches, such as the VIENNA rectifier, for applications
above 6kW. Possible solutions for reducing the cost of the front-end PFC were presented, such
as the single- and two-switch DCM boost rectifiers. Such PFC configurations were also
interleaved to help reduce the combined size of boost and filter inductors.
As an alternative solution, the PFC and DC/DC converters can be integrated into one power
stage to reduce the front-end converter cost, as described in previous work [71]-[74]. These
solutions, except for one [71], cannot handle very high power levels because they use single-
switch topologies, such as the three-phase cuk, zeta and flyback converters. Moreover, these
topologies are well known for increasing the voltage stress across the switch. In one solution
[71], a phase-shift full-bridge topology was used to implement a single-stage converter. Despite
the need for an auxiliary transformer winding that operates as a magnetic switch to control the
PFC function, the overall system [71] is simpler than those discussed in the previous chapters.
However, the single-stage converter previously presented requires the power switches to
withstand the total intermediate bus voltage stress [71]. This problem turns out to be critical,
since the intermediate bus voltage in single-stage converters fluctuates according to input voltage
and load variations. Therefore, power switches with high voltage ratings are required if the
220V.
To overcome this problem, the work presented in this chapter proposes novel three-phase and
three-level single-stage AC/DC converters that are able to achieve low harmonic distortion and
reduced parts count, as well as to reduce cost when compared to those solutions discussed in the
previous chapters. The proposed three-level converters discussed in this chapter decrease the
voltage stress across the power switches, achieve ZVS without auxiliary circuits, and require
This chapter presents a comparison of different approaches for single-stage three-phase AC/DC
power conversion. Different aspects are compared throughout the chapter, such as intermediate
bus voltage stress, harmonic distortion, efficiency, interleaved operation, and ability to constrain
psophometric noise.
In a two-stage approach, the rectifier and DC/DC converter operate independently from each
other because the intermediate bus capacitor is large enough to decouple the operation and the
control of both stages, as illustrated in Fig. 4-1. The advantage of the two-stage approach is that
the PFC circuit provides a regulated intermediate bus voltage (800V in this case), which
facilitates the design optimization of both converters with respect to efficiency. Since in a two-
stage front-end converter the design of both stages can be more easily optimized, the overall
reduction.
S5
S2
S6 S3
S4
D6
PFC DC/DC
Controller Controller
The single-stage approach can reduce cost because of the reduced number of switches and
controllers needed to shape the input current and to regulate the DC output voltage [75] [76].
Although the cost of single-stage front-end converters is lower than that of a two-stage approach,
the intermediate bus voltage can no longer e regulated because the controller in the circuit is used
to regulate only the DC output voltage. Consequently, the variation of the intermediate bus
It is envisioned that with increased power levels for mainframe computers and servers, lower-
cost solutions will become more in-demand in the future. Trying to address the ever-increasing
requirement for lower-cost applications, this chapter addresses the functional integration of PFC
The main objective for integrating the power stages is to eliminate both switches S5 and S6 and
diodes D5 and D6 from the PFC stage shown in Fig. 4-1. There are very few steps necessary for
the integration of the power stages. The first step is to eliminate the devices S5, S6, D5 and D6
from the PFC stage shown in Fig. 4-1. The second step is to connect the output of the three-phase
input bridge to the positive and negative DC rails, as illustrated in Fig. 4-2(a). As can be seen in
the same figure, the PFC function can no longer be performed because of the way the artificial
neutral is connected to the power stage. In fact, the input section would operate as a conventional
bridge-type rectifier supplying a capacitive filter, which would require huge input peak currents
to charge the bus capacitors. To overcome this problem, the third step in the synthesis of the
single-stage converter is to move the artificial neutral point connection to point B. The result of
reconnecting the artificial neutral point is shown in Fig. 4-2(b). The four switches that remain in
the circuit operate with phase-shift control to transfer power to the output, as well as to store
energy in the input inductors designed to operate in DCM. Fig. 4-3 shows the main operating
This section presents the operating principle of the proposed single-stage TL-PS converter. The
converter shown in Fig. 4-2(b) contains four power switches S1 to S4, three input boost inductors
La to Lc, two intermediate bus capacitors Cb1 and Cb2, two clamping diodes Dc1 and Dc2, a
clamping capacitor Cf for ZVS operation of the outer switches, two output rectifiers Dr1 and Dr2,
an output filter consisting of Lo and Co, and a transformer that isolates the output voltage from
the transformer to increase the load range with soft-switching operation. The capacitors Ca, Cb
and Cc provide an instantaneous three-phase voltage source and an artificial neutral point
capacitors used to provide the artificial neutral point connection are not extra components, since
S1
S2
B
S3
S4
(a)
Lr
S1 Dr1
Lo
La
Cb1 Dc1
Cf Co
Lb S2
A B
Lc
Cb2 Dc2 S3
Ca Cb Cc
Dr2
S4
(b)
Fig. 4-2. Synthesis of TL-PS single-stage AC/DC converter: (a) eliminating the PFC cell, and (b)
reconnecting the artificial neutral point.
S2 S3 S2
vab
iLr
ic
ia
ton
ib
to t1 t5 t6
t2 t3 t4
To explain the operating stages, the components in the converter are considered ideal, the
intermediate bus capacitors are constant voltage sources, and the AC input voltages are constant
within one switching cycle. The capacitors Ca, Cb and Cc and the input filter will be ignored for
simplicity of explanation. To achieve low harmonic distortion in the input currents, the boost
inductors must operate in DCM. Nevertheless, the DC side of the three-level converter may be
operated either in CCM or DCM. In the description that follows, only CCM operation is
The proposed AC/DC converter is controlled by conventional phase-shift control. All switches
operate with nearly 50% duty cycle, while a dead time is required to avoid shoot-through of the
power switches. As can be seen in the diagram shown in Fig. 4-2(c), the outer switches S1 and S4
are 180o apart. The inner switches S2 and S3 are also complementary, while phase-shifting the
the modulation strategy applied to the conventional full-bridge phase-shift converter [77] [78],
the leading leg in the three-level structure is associated with the outer switches, while the lagging
Stage 1 (to, t1), Fig. 4-4(a): During this stage, the switches S1 and S2 are turned on. The currents
ia and ic rise linearly while ib decreases, as illustrated by the waveforms in Fig. 4-3. Power is
Stage 2 (t1, t2), Fig. 4-4(b): At instant t1, S1 is turned off and C1 is charged towards the voltage
across the capacitor Cb1. Simultaneously, C4 is discharged towards zero through the clamping
capacitor Cf. This stage ends when the diode Dc1 clamps the voltage across the capacitance C1 to
the voltage across Cb1. At the same time, the voltage across C4 reaches zero.
Stage 3 (t2, t3), Fig. 4-4(c): This is the freewheeling stage of the output current that starts when
the voltage across C1 reaches the voltage across Cb1, turning on the diode Dc1. The input currents
Stage 4 (t3, t4), Fig. 4-4(d): At instant t3, S2 is turned off and the resonant inductance placed in
the primary side for the purpose of soft switching resonates with the capacitances C2 and C3. The
voltage across C2 rises to half of the intermediate bus voltage, while the voltage across C3
decreases to zero to complete the stage. The currents ia and ic continue decreasing during this
stage.
Stage 5 (t4, t5), Fig. 4-4(e): During this interval, the anti-parallel diodes of S3 and S4 conduct
both the primary current of the transformer and the remaining levels of input AC currents. The
current in phase b starts building up because the conduction of the intrinsic diodes of S3 and S4
Stage 6 (t5, t6), Fig. 4-4(f): During this stage, S3 and S4 are turned on. Currents through La and Lc
decrease to zero, and the current through Lb increases linearly. Power is delivered to the output
side during this interval. After instant t6, another half operating cycle will be repeated for
This section analyzes the operation of the proposed converter to devise design guidelines. The
discussion is divided in two parts: (1) AC-side and (2) DC-side analyses. According to the
operating stages, the boost inductor currents are independent from each other because the neutral
point is connected to the switching power stage. As a result, the analysis performed in one phase
A. AC-Side Analysis
For this analysis, the three-phase input voltages are balanced, while the voltage across phase a is
taken as the reference voltage for the three-phase system. A conceptual representation of the
current in the boost inductor connected to phase a is shown in Fig. 4-5 for half of a line cycle.
During the intervals 0<θ<φcr and (π-φcr)<θ<π, the boost inductor current in phase a is completely
reset during the freewheeling stage (stage 3 shown in Fig. 4-4(c)). However, for the remaining
interval φcr<θ<(π-φcr) of the half-line cycle, the time duration of the freewheeling interval is not
sufficient to completely reset the boost inductor current in phase a. This is the reason why the
boost current waveforms assume different slopes in the illustration shown in Fig. 4-5.
Dr1 Lo Ro Dr1 Lo Ro
N2 N2
Co Co
N2 Dr2 N2 Dr2
Dr1 Lo Ro Dr1 Lo Ro
N2 N2
Co Co
N2 Dr2 N2 Dr2
Dr1 Lo Ro Dr1 Lo Ro
N2 N2
Co Co
N2 Dr2 N2 Dr2
The critical angle φcr in which the current in phase a is completely reset during the freewheeling
interval is given by
where Vbus is the total intermediate bus voltage (sum of the voltages across Cb1 and Cb2), Vpk is
the peak line-to-neutral input voltage, and D is the operating duty cycle.
[A]
π θ
φcr φcr
Fig. 4-5. Illustration of the boost inductor current waveform for half of a line period.
The average input phase current is obtained from the high-frequency waveform schematically
shown in Fig. 4-5. For the intervals 0<θ<φcr and (π-φcr)<θ<π, the average input current is given
by
Vbus sin (θ )
(4-2) iin1 (θ ) = D 2 ,
V
2 bus − 2 sin (θ ) L f s
V pk
During the remaining interval φcr<θ<(π-φcr), the average input current is given by
3 φcr π 2
(4-4) P= ∫ V pk sin (θ ) iin1 (θ ) dθ + ∫ V pk sin (θ ) iin2 (θ ) dθ .
π 2 0 Φ cr
The minimum bus voltage condition for proper operation of the proposed converter can be easily
derived from the freewheeling stage shown in Fig. 4-4(c). As can be verified from that figure, at
the peak voltage across phase a, the total intermediate bus voltage must be Vbus≥2Vpk in order to
provide reset voltage for the input current during the freewheeling stage (the reset mentioned
here is partial and does not necessarily need to bring the boost inductor current to zero). For the
polarity of the input voltages assumed in Fig. 4-4, when the total reset of the boost inductor
current flowing through phase a is not achieved during the freewheeling stage, it must then be
Fig. 4-6 shows the normalized intermediate bus voltage as a function of the output power. As can
be seen, the PFC section of the proposed converter behaves as a current source, since the input
inductors operate in DCM. The minimum intermediate bus voltage is twice the input peak line-
Vbus
M =
V pk
(4-5) .
Po L f s
Pnorm =
V pk 2
3
D=0.4
D=0.1 D=0.3
2.5
D=0.2
2
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Normalized output power
Fig. 4-6. Normalized bus voltage gain.
B. DC-Side Analysis
Depending on the output load condition, the DC side of the proposed converter operates either in
D 4
(4-6) Vo = Vbus − 2 I o Lr f s ,
nt nt
where nt is the transformer turns ratio, Io is the output load current, and Lr is the total resonant
inductance connected in series with the primary side of the transformer to provide energy for soft
switching. Equation (4-6) shows that Lr reduces the output voltage according to the load current.
The second term in (4-6) is a factor common to all converters that rely on the energy stored in
the resonant inductor to achieve soft switching for the lagging switches [13].
When the output inductor of the DC side operates in DCM, the output voltage is given by
The results of the DC-side analysis are plotted in Fig. 4-7. The boundary line between DCM and
CCM shown in Fig. 4-7 is associated with the operating mode of the output inductor. As can be
observed in the CCM region, there is a drop in the output voltage caused by the resonant
inductance Lr. As the resonant inductance becomes larger, the circulating energy and the output
voltage drop also increase in the proposed converter. The normalized variables used to plot the
0.5
Output voltage gain (Mout)
D=0.5
0.4
D=0.4
0.3
D=0.3
0.2
D=0.2
0.1
CCM/DCM D=0.1
boundary
0
0 0.05 0.1 0.15 0.2
Normalized output current (Io-norm)
nt Vo
M out =
Vbus
(4-8) ,
n L f
I o − norm = Io t o s
Vbus
inductance ratio nt 2 Lo Lr = 8 . As the inductance ratio increases, the region in which the output
section of the converter operates in CCM also increases. However, a larger CCM operating range
in the output section of the proposed converter will increase the intermediate bus voltage, as
This section describes a simplified design procedure based on the theoretical analysis presented
A. Specifications
The proposed single-stage TL-PS converter is designed to operate with an input line-to-neutral
voltage of 170V to 265V RMS, 3kW of output power, and 48V of DC output voltage while
To optimize the design of the proposed converter, the AC side must operate as close as possible
to the DCM/CCM boundary condition, which is associated with the minimum intermediate bus
voltage gain M=2. As already described, for proper operation of the proposed converter, the
intermediate bus voltage must be two times greater than the peak input phase voltage. Under
such an assumption, if the bus voltage gain (Vbus/Vpk) is set to 2.1, the input currents will be
forced to operate in DCM. Choosing a minimum voltage gain of 2.1, the intermediate bus
voltage will be 504V at low-line input phase voltage (170V) and full load.
The duty cycle is chosen to provide output voltage regulation at full load and low-line input
voltage. The maximum duty cycle D is chosen to be 0.45, thus providing compensation for the
dead time between the gate signals of the power switches. Choosing a large duty also increases
D. Boost Inductance
The power delivered to the output is limited by the boost inductances. According to (4-4) and the
results from the previous analysis, the power delivered to the output can be represented as a
(4-9) ( )
P = f Vbus ,V pk , D, L, f s .
At low-line input voltage and full-load condition, all parameters in (4-9) are known, except that
the boost inductance must be determined. Solving (4-4) for the boost inductance and taking into
account the design point already established above (170V of input phase voltage, 504V of
intermediate bus voltage, 0.45 of duty cycle, and 100kHz of switching frequency), the resulting
The second term in (4-10) is the well-known duty-cycle loss. Increasing this term means
obtaining ZVS at lighter loads. However, increasing the duty-cycle loss also increases the
circulating energy in the converter. Choosing a duty-cycle loss equal to 0.07 yields nt=4.
140
Po=3kW
Boost inductance (µH)
120
Dmax=0.45
Vbus-min=504V
100 Vin-min=170V
80
60
40
40 50 60 70 80 90 100
Frequency (kHz)
Fig. 4-8. Boost inductance as a function of the switching frequency for DCM operation.
The resonant inductance is obtained from the duty-cycle loss term defined in (4-10), as follows:
nt Dloss Vbus
(4-11) Lr = .
4 Io fs
In this case, the full-load output current is considered in the calculation in order to guarantee
regulation of the output voltage. Substituting all pertinent values into (4-11), the resonant
inductance is 2.83µH.
The impact of the inductance ratio on the maximum intermediate bus voltage is shown in Fig.
4-9. As can be seen, the intermediate bus voltage increases with the inductance ratio. The
inductance ratio, nt 2 Lo Lr , influences the current ripple in the output inductor, or more
specifically, influences the extent to which the output section is designed to operate in CCM. As
the inductance ratio increases, the region in which the output inductor operates in CCM also
1300
n=4
1250
Vin=265V (RMS)
Bus voltage (V)
1200
1150
1100
1050
1000
4 6 8 10 12
Inductance ratio (nt2Lo/Lr)
Fig. 4-9. Calculated maximum intermediate bus voltage stress as a function of inductance ratio.
To obtain the THD of the proposed converter, the Fourier analysis can be applied to the
instantaneous average boost inductor current given in (4-2) and (4-3). It is important to mention
that the instantaneous average boost inductor current given in (4-2) and (4-3) includes the zero-
As a result, a third harmonic circulates in the power lines. However, as mentioned before, the AC
capacitors are used to provide an artificial neutral point connection in order to eliminate the
The THD of the instantaneous average boost inductor current, including the third harmonic, is
plotted in Fig. 4-10(a). As can be verified, the distortion is high when the third harmonic is not
eliminated from the line current. However, Fig. 4-10(b) shows that the distortion is very low
when the zero-sequence-order harmonics are eliminated from the power source by using the
artificial neutral point connection created by the AC capacitors. As shown in Fig. 4-10, the THD
of the input line currents depends upon the duty cycle. This result makes sense because when the
duty cycle increases, the freewheeling time is reduced, and consequently the faster slope of the
reset seen in Fig. 4-3 dominates the reset of the boost inductors, thus reducing the THD. Fig.
4-10(b) shows a comparison between the THD produced by the proposed converter and that
generated by the three-phase single-switch DCM boost rectifier [42]. For the same intermediate
bus and input voltages, the THD generated by the proposed converter is much lower than that
which is generated by the three-phase single-switch DCM boost rectifier (observe that a log scale
was used to plot the THD without the zero-sequence order harmonics).
This section presents the experimental results obtained from the proposed TL-PS AC/DC
converter. The implemented switching power stage used IXFN44N80 MOSFETs, RUR30120
line rectifiers and clamping diodes, and HFA120MD40D output rectifiers. Each bus capacitance
capacitor Cf and the AC capacitors Ca, Cb and Cc were implemented with 2µF polypropylene-
type capacitors. The output filter inductance was 3µH, while the output filter capacitor was
2x4700µF/100V.
50 100
D=0.1 D=0.1
THD with third harmonic (%)
D=0.2
10
20
10
(b)
0 1
2 2.5 3 3.5 4 4.5 2 2.5 3 3.5 4 4.5
Bus voltage gain, Vbus/Vpk Bus voltage gain, Vbus /Vpk
(a) (b)
Fig. 4-10. THD: (a) with third harmonic, and (b) without third harmonic.
Fig. 4-11(a) shows the measured input phase voltage and the input current of the proposed
converter at full load. The THD of the current waveform illustrated in Fig. 4-11(a) is only 4.8%.
The fifth harmonic is still dominant in the input current, accounting for an individual distortion
of 4.3%. Fig. 4-11(b) shows the voltage vab and the primary transformer current as indicated in
Fig. 4-2(b). These waveforms are similar to the classic waveforms produced by the DC/DC
converters with phase-shift control and ZVS operation. Fig. 4-11(c) shows the voltages and
currents through switches S2 (inner switch) and S4 (outer switch). As can be seen, the outer and
The intermediate bus voltage measured at three different input voltages is shown in Fig. 4-12(a).
The bus voltage depends on the input voltage as well as on the output load variations. For a
is a region in the plots where the intermediate bus voltage remains approximately constant. In
that region, the DC side of the proposed converter starts operating in DCM. This operating mode
helps to limit the increase of the bus voltage. Therefore, by reducing the output inductance, the
power level at which the DC side starts operating in DCM decreases and the intermediate bus is
limited to lower voltages. However, reducing the output inductance incurs more current stress in
the primary switches and more current ripple in the output section. According to Fig. 4-9,
reducing the output inductance is equivalent to reducing the inductance ratio nt 2 Lo Lr , which
vin
vab
iin
iLr
(a) (b)
vS4
iS4
vS2 iS2
(c)
Fig. 4-11. Experimental waveforms: (a) input voltage (100V/div) and filtered input current (5A/div),
(b) voltage vab (250V/div) and primary current (10A/div), and (c) ZVS waveforms (voltages: 200V/div
and currents: 20A/div).
The harmonic distortion is always lower than 9%, even at high-line input voltage. It is also
noteworthy that the experimental THD is higher than the theoretical distortion shown in Fig.
4-10(b). The reason for this difference can be attributed to the fact that the AC capacitors used to
provide the artificial neutral connection do not result in ideal voltage sources (designed as those
that are free of harmonics), and consequently the THD of the input currents is higher than that
The measured converter efficiency at full load, 100kHz, and 220V input line-to-neutral voltage
is shown in Fig. 4-12(c). Despite the ZVS operation provided by the converter, the efficiency
is rather low for the power level under discussion. The main reason for the low efficiency is the
high input current ripple that occurs due to the DCM operation of the boost inductors. The
circulating energy that is seen in the current waveform iLr and the turn-off loss of the power
A solution for reducing the switching loss is to decrease the switching frequency. Fig. 4-12(c)
shows the improvement in efficiency gained by cutting back the switching frequency to 50kHz.
At 50kHz, the converter parameters are L=110µH, Lr=11.3µH, and Lo=5.7µH. All the power
devices were the same as those used for the design at 100kHz in order to retain the same level of
conduction loss in the circuit. As can be verified, the efficiency at 50kHz is significantly
improved, which proves that the turn-off loss plays an important role in converter efficiency. The
turn-off loss is rather high because the switches carry both the boost inductor currents from the
AC side and the DC output current reflected to the primary side of the transformer.
1100 10
9 Vi = 260V
1000 Vi = 260V
8
900 7
Bus voltage (V)
800 6 Vi = 220V
THD (%)
5
700 4
Vi = 180V Vi = 180V
Vi = 220V
600 3
2
500
1
400 0
180 680 1180 1680 2180 2680 3180 180 680 1180 1680 2180 2680 3180
Output Power (W) Output Power (W)
(a) (b)
89
88
87 fs=50kHz
Efficiency (%)
86
85
84
83
82 fs=100kHz
81
80
170 190 210 230 250 270
Input line-to-neutral voltage (V)
(c)
Fig. 4-12. Experimental results: (a) intermediate bus voltage, (b) THD, and (c) converter efficiency
measured for both designs.
The reason for the reduced efficiency of the TL-PS single-stage AC/DC converter can be better
understood by referring to Fig. 4-13. In order to transfer power to the output and to store energy
in the input inductors, either the upper or the lower switches must be on. As can be seen, the
rectified input currents flow through both switches at the same time, which increases the circuit
conduction loss. Moreover, since the input currents are discontinuous, the switches are turned off
at the peak of the rectifier input currents during every switching cycle, thus increasing the
S1
Lr
S2
A
B
S3
S4
The three-level DC/DC converter depicted in Fig. 4-1 was based on the neutral-point-clamped
(NPC) three-level inverter [79], which also served as the starting point for the single-stage TL-
PS AC/DC converter discussed in the previous section. However, other multilevel topologies can
also be used to realize different three-level DC/DC converters [80]-[82]. For instance, one
topological variation for the DC/DC converter is illustrated in Fig. 4-14(a). This topology uses
stacked switches to achieve a multi-level arrangement that is simpler than the NPC structure. The
control of the converter illustrated in Fig. 4-14(a) is based on the asymmetrical PWM control,
C2 S4
vab
(a) (b)
Fig. 4-14. Three-level asymmetrical DC/DC converter: (a) topology and (b) waveforms.
The topology shown in Fig. 4-14(a) can easily be combined with a rectifier bridge to form an
AC/DC converter similar to the one described in the previous section of this chapter. The starting
point is the two-stage front-end converter based on the two-switch three-level DCM boost
4-15(a). The synthesis of the single-stage converter is obtained by removing the switches and
diodes of the PFC stage, while connecting the positive and negative outputs of the bridge
In the proposed circuit, the four switches are connected in a stacked three-level structure, while
an asymmetrical PWM is used to control the DC output voltage. The capacitor Cb is used to
provide voltage balance across the transformer because of the asymmetrical duty ratio imposed
by the control. The current source in parallel with the transformer is fictitious and represents the
DC bias of the magnetizing current caused by the asymmetrical PWM. The resonant inductance
the proposed converter. The input boost inductors must operate in DCM to guarantee low THD
for the input currents, while the AC capacitors are incorporated by the input filter and provide an
artificial neutral connection point for the switching power stage. The bus capacitors C1 and C2
provide energy storage as well as resetting voltage to the input boost inductors. Either the inner
or outer switches are turned on and off simultaneously in order to transfer power to the output
section. The following paragraphs describe the operating stages of the proposed converter.
S5
S2
S6 S3
S4
D6
(a)
S1
Tr Lo
A Lr
Lb Co
S2
Input Filter
B
S3
Cf Cb
S4
(b)
Fig. 4-15. Synthesis process: (a) two-stage approach, and (b) TL-AS single-stage AC/DC converter.
iLr
vab
ic
ia
ib
to t1 t2 t3 t4
Stage 1 (to, t1), Fig. 4-17(a): During this stage, S2 and S3 are on in order to store energy in the
input inductors, while the blocking capacitor is discharged across the transformer and output
filter.
Stage 2 (t1, t2) Fig. 4-17 (b): In this stage, the switches S2 and S3 are turned off. The rectified
boost inductor currents and the current in Lr charge the intrinsic capacitances of S2 and S3, as
well as discharging the capacitances of S1 and S4. When the voltages across S1 and S4 reach zero,
the rectified input currents and iLr are diverted into the intrinsic diodes of the switches S1 and S4,
providing the condition for zero-voltage turn-on for S1 and S4. During this stage, the resonant
inductor current will reverse its polarity because the total bus voltage is greater than the voltage
across the DC blocking capacitor Cb. Because S1 and S4 operate for longer times than S2 and S3,
the DC bias of the transformer magnetizing current builds up according to the polarity shown in
Fig. 4-17. Therefore, this stage ends when the primary current of the transformer reaches the
transformer.
Stage 3 (t2, t3), Fig. 4-17 (c): During this stage, power is transferred from the bus capacitors C1
and C2 to the output, while the input inductors are reset by the difference between half of the
intermediate bus voltage and the input phase voltage source connected to the boost inductor.
Stage 4 (t3, t4), Fig. 4-17 (d): When the switches S1 and S4 are turned off, the primary current
will charge the intrinsic capacitances of S1 and S4, as well as discharging the capacitances of S2
and S3. Once this transition has been finalized, the remaining primary current will circulate
through the intrinsic diodes of S2 and S3, which provides the condition for zero-voltage turn-on.
During this stage, the primary current will reverse its polarity until it reaches the sum of the
reflected output current and the DC magnetizing current. Once this condition has been reached,
This circuit variation can reduce the conduction loss because the input currents, after being
rectified, circulate through only one switch at a time, as opposed to the circuit discussed in the
previous section.
This section provides the analytical results obtained from the proposed topological variation. The
discussion is also divided into AC-side and DC-side analyses. The neutral point connected to the
switching power stage decouples the input phase currents from each other. As a result, the
analysis performed for one phase can be extended to the other phases as well.
(a) (b)
C1 + S1
Tr Lo C1 + S1
Tr Lo
a Lr + a Lr +
La Co La Co
+ S2 + S2
Lb Lb
+ +
+ iLr + iLr
Lc S3 b Lc S3
+ + b
Cb Cb
+ +
C2 S4 C2 S4
(c) (d)
A. AC-Side Analysis
For this analysis, the three-phase input voltages are considered balanced voltages, while the
voltage across phase a is used as the reference voltage. The peak boost inductor currents can be
D
0 0
I a − pk L f s v a
D
(4-12) I b − pk = 0 0 vb ,
L fs
I c − pk
0 D vc
0
L f s
where va, vb and vc are the three-phase input voltages given by (2-1) (see page 21).
total intermediate bus voltage and the instantaneous input phase voltage applied across the boost
inductor. For each of the input phases, the time needed to reset the boost inductor is
D
2V pk sin (θ )
da
t
( )
Vbus − 2V pk sin (θ ) f s
t = 2V sin θ + π D
(4-13)
db pk
3 π
.
t dc Vbus − 2V pk sin θ + f s
3
π D
2V pk cosθ +
6 π
Vbus − 2V pk cosθ + f s
6
After determining the peak boost inductor currents and the time needed to reset each boost
inductor, the instantaneous average boost inductor currents over a half-line period can be
determined as follows:
Vbus
D V pk sin (θ )
2
a
i
(
2 L f s Vbus − 2V pk sin (θ ) )
i = D 2 V sin θ + π V bus
(4-14)
b .
pk
3 π
ic 2 L f s − Vbus + 2V pk sin θ +
3
2 π V bus
D V pk cosθ +
6 π
2 L f s Vbus − 2V pk cosθ +
6
The previous result can be used to determine the amount of power transferred to the output as a
where
Vbus
(4-16) M = , and
V pk
Po L f s
(4-17) Pnorm = .
V pk 2
The voltage gain (Vbus/Vpk) is plotted in Fig. 4-18 as a function of the normalized output power,
using D as a running parameter. Notice that because of the slope of the voltage gain, the AC side
4
D=0.4
D=0.3
Normalized bus voltage
3.5
D=0.2
3
D=0.1
2.5 CCM/DCM
Boundary
2
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Normalized output power
Fig. 4-18. Normalized bus voltage gain for the TL-AS AC/DC converter.
which yields:
2
(4-18) M cr = .
1− D
Therefore, to guarantee operation of the topological variation in DCM, one has to ensure that
The DC blocking capacitor is used to balance the voltage across the transformer. The voltage
across Cb depends upon both duty cycle and intermediate bus voltage. The voltage ripple across
Cb is controlled by its capacitance value and the amount of current that flows in the primary side.
In order to balance the voltage across the transformer, the DC blocking capacitor voltage will
(4-19) Vc = (1 − D )Vbus .
Although the DC blocking capacitor provides balance to the voltage across the transformer, the
asymmetrical duty cycle operation results in an average magnetizing current through the
transformer. The level of DC bias depends upon the load current and duty ratio, as shown below.
(4-20) I DC −mag =
(1 − 2 D ) I
o.
nt
B. DC-Side Analysis
The DC side of the TL-AS single-stage converter can operate either in DCM or CCM, depending
upon the load. For CCM operation, the output voltage is given by
Equation (4-21) shows how the voltage drop across Lr reduces the output voltage as a function of
The output side can also operate in DCM, which requires a new set of equations to describe the
output voltage. Fig. 4-19(a) shows the current waveform through Lo when the DC output
operates in DCM. There are three sub-intervals during the DCM operation of the DC output side.
As can be seen, these sub-intervals are associated with different equivalent circuits. For
simplicity, these equivalent circuits are obtained by neglecting the magnetizing current of the
transformer. The average DC output current can be determined by the following relationship:
(4-22) Io =
1 D Ts I pk
+
( )
I pk + I q ∆t t q I q
+ .
Ts 2 2 2
All the variables in (4-22) can be determined from the equivalent circuits shown in Fig. 4-19. For
instance, during the magnetizing interval of Lo, switches S2 and S3 are on. From the circuit
shown in Fig. 4-19(b), one can conclude that the peak current through inductor Lo is given by:
nt D
I pk = (Vc − nt Vo )
(4-23)
(Lr )
+ nt 2 Lo fs
.
Once the interval DTs is finished, the switches S2 and S3 will be turned off. For simplicity of
analysis, the transition interval needed to charge and discharge the intrinsic capacitances of the
power switches will be neglected. Therefore, the circuit shown in Fig. 4-19(c) is related to the
interval needed for the resonant inductor to reverse its current polarity. Meanwhile, the output
2 I pk Lo Lr
∆t =
Lr Vo + nt Lo (Vbus − Vc )
(4-24) .
2Vo I pk Lr
I q = I pk −
Lr Vo + nt Lo (Vbus − Vc )
The equivalent circuit shown in Fig. 4-19(d) determines the behavior of the current in the output
inductor when the current through Lr reflected to the output reaches Iq. From the circuit shown in
Fig. 4-19(d), the time interval needed to finish resetting the output inductor is given by:
(4-25) (
t q = Lr + nt 2 Lo )n (n V Iq
− Vbus + Vc )
.
t t o
Equations (4-23), (4-24), (4-25) and (4-19) can be substituted into (4-22) to yield the DC output
voltage. Because the resulting expression for the DC output voltage is too large to fit on the
page, the result is given in terms of the variables that affect the DC output voltage. The function
(4-26) (
Vo = g I o , Lo , Lr , f s , nt , D,V pk ,Vbus . )
This section describes a simplified design procedure for the single-stage TL-AS converter, and
provides an example based on the theoretical analysis presented in the previous section.
Vc/nt + +
- - Vo
D Ts ∆t tq t
(b)
(a)
Lr Lo Lr/nt2 Lo
Vb-Vc + +
Vo
+ +
- - - (Vb-Vc)/nt - Vo
(c)
(d)
Fig. 4-19. DC-side DCM operation: (a) inductor current through Lo, (b) magnetizing stage for Lo, (c)
resetting stage for Lo and reversing current polarity in Lr, and (d) final resetting stage for Lo.
A. Specifications
The proposed single-stage TL-AS converter is designed to operate within an input line-to-neutral
voltage variation of 170V to 265V, while providing 3kW at 48V to the output, and switching the
B. Duty Cycle
According to the characteristics of the single-stage TL-AS converter shown in Fig. 4-18, for a
given normalized output power the intermediate bus voltage increases as the duty cycle
increases. From this standpoint, the duty cycle cannot be too high. On the other hand, choosing a
small duty cycle increases the current stresses in both the AC and DC sides. As a result of this
trade off, the duty cycle for the design of the single-stage TL-AS converter has been chosen to be
0.25. This duty cycle is chosen to provide output voltage regulation at full load and low-line
input voltage.
near the DCM/CCM boundary condition. The voltage gain at the DCM/CCM boundary is given
by (4-18). Using the chosen duty cycle (D=0.25), the critical voltage gain at the DCM/CCM
boundary is 2.67, which results in an intermediate bus voltage of 641V at low-line input voltage
(170V).
nt Vo 4I L f
(4-27) = 2 D (1 − D ) − o r s .
Vbus nt Vbus
The second term in (4-27) represents the duty-cycle loss in the DC side caused by the circulation
of energy through the resonant inductor Lr. Increasing this term means increasing the load range
in which the proposed converter operates with ZVS. On the other hand, increasing the duty-cycle
loss also increases the circulating energy in the converter. Therefore, a duty-cycle loss of 20%
with respect to the first term in (4-27) represents a good trade off between the ZVS range and the
The transformer turns ratio can then be calculated from (4-27) using the converter specifications
at full load and low-line input voltage. At this operating point, the bus voltage is 641V, as
The resonant inductance is obtained from the duty-cycle loss term, defined as follows:
The calculation is performed at full load to guarantee regulation of the output voltage.
Substituting all the pertinent values into (4-28) results in the resonant inductance Lr=15.41µH.
The trade offs in determining the output inductance are similar to those discussed for the single-
stage TL-PS converter presented in the beginning of this chapter. Increasing the output
inductance also increases the maximum intermediate bus voltage of the proposed single-stage
TL-AS converter. An inductance ratio of nt 2 Lo Lr = 8 has also been used in the design of the
E. Boost Inductance
The power delivered to the output is limited by the input boost inductances. According to (4-15),
the normalized power delivered to the output at low-line input voltage and full load is Pon=0.142.
The boost inductance can be obtained by de-normalizing (4-17), which results in 55µH.
The use of AC capacitors to provide the artificial neutral point connection changes the AC input
to the intermediate bus voltage converter gain. For this reason, when the boost inductance
determined above is connected in the presence of the AC capacitors, it is observed that the boost
inductor operates in a deeper DCM condition as compared to the theoretical prediction. Taking
into account the design point at low-line input voltage and full load, the boost inductance has to
be adjusted to 72µH in order to make the proposed converter operate as predicted in the design.
The boost inductance at different switching frequencies can be determined from Fig. 4-20.
60
40
20
40 50 60 70 80 90 100
Switching frequency (kHz)
Fig. 4-20. Boost inductance as a function of the switching frequency for DCM operation.
This section presents experimental results obtained from the single-stage TL-AS converter, and
provides a comparison with the single-stage TL-PS converter discussed at the beginning of this
chapter. Table 4-1 summarizes the various components used in the implementation of both
converters. As can be seen, the major difference between them is that the transformer of the
single-stage TL-AS converter requires more turns to account for the DC magnetizing current that
results from the asymmetrical operation. The transformer core size, however, was the same for
Fig. 4-21(a) and Fig. 4-21(b) show the drain voltages and gate signals for switches S1 and S2,
respectively. As can be observed, both switches operate under ZVS, since the drain-to-source
voltages reach zero before the gate signals are applied. These results have been obtained at 3kW
of output power and 180V of line-to-neutral input voltage. Since the total intermediate bus
voltages. The asymmetrical characteristic of the converter can be seen in the figure, while the
average current through the primary side of the transformer is zero. Fig. 4-21(d) illustrates the
input current in one of the phases after the switching ripple has been filtered for three different
input phase voltages. As observed, the harmonic distortion of the input current is dependent upon
Fig. 4-22(a) shows the intermediate bus voltage measured from the TL-AS converter for three
different input phase voltages. The intermediate bus voltage increases as the output power
decreases. In the same way as described before for the TL-PS converter, the explanation for this
behavior is based on the power balance and the operating mode of the output filter inductor. For
instance, suppose that the output inductor is operated in CCM and that the voltage drop across
the resonant inductor can be ignored. Under these assumptions, the DC output voltage is simply
a function of the duty cycle and intermediate bus voltage. If the load decreases, the duty cycle
must also decrease to store less energy in the DCM boost inductors. Since the DC output voltage
decreasing the duty cycle has to be compensated for by increasing the intermediate bus voltage
in order to maintain the DC output voltage regulation. This mechanism explains why the
intermediate bus voltage increases as the output power decreases. Fig. 4-22(a) also shows a
region at light load in which the intermediate bus voltage no longer increases. In that region, the
output inductor starts operating in DCM. As can be seen, the intermediate bus voltage of the TL-
AS converter reaches the same level at light load as the intermediate bus voltage of the TL-PS
As illustrated in Fig. 4-22(b), there is voltage stress across the DC blocking capacitor. For this
reason, it is required that either polypropylene or metallic film capacitors be used in order to
withstand the voltage stress, as well as to withstand the primary RMS current, which is 15A at
Fig. 4-22(c) illustrates the efficiency comparison between three converters: (1) the two-stage
front-end converter studied in chapter 3, (2) the single-stage TL-PS converter presented at the
beginning of this chapter, and (3) the single-stage TL-AS topology under discussion. As
predicted, the single-stage TL-AS converter presents improved efficiency when compared
against the singe-stage TL-PS converter. The same figure also shows the trade off between
performance and cost. The two-stage approach presents the best efficiency of all cases under
(a) (b)
220V
260V
10A/div 5A/div
(c) (d)
Fig. 4-21. Experimental results: (a) drain voltage and gate signal for S1 at 3kW and Vin=180V, (b)
drain voltage and gate signal for S2 at 3kW and Vin=180V, (c) primary current at 3kW for three input
voltages, and (d) filtered input current at 3kW for three input voltages.
The harmonic distortion comparison between the three cases is illustrated in Fig. 4-22(c).
Despite presenting an efficiency improvement over the single-stage TL-PS converter, the TL-AS
topology does not perform well in terms of THD. The lowest THD is obtained from the TL-PS
converter, while the THD of the two-stage approach falls between the two single-stage
converters.
900
900 Vin = 260V
850 800
Vin = 220V
800 700
750
600
700 Vin = 180V Vin = 220V
Vin = 180V 500
650
600 400
0 500 1000 1500 2000 2500 3000 3500 0 1000 2000 3000 4000
Output power (W) Output power (W)
(a) (b)
94 16
93 14
92 12
Efficiency (%)
Two-stage
91 THD (%) 10
TL-AS
90 8
TL-AS Two-stage
89 6
88 4
TL-PS
87 2
TL-PS
86 0
160 180 200 220 240 260 160 180 200 220 240 260
Line-to-neutral input voltage (V) Line-to-neutral input voltage (V)
(c) (d)
Fig. 4-22. Results and comparisons: (a) intermediate bus voltage stress,(b) DC blocking capacitor,
(c) efficiency comparison, and (d) THD comparison.
Both single-stage three-level AC/DC converters operate under DCM. As a result, the input
current ripple is significantly higher, as compared to boost-type CCM converters. This ends up
increasing the size and volume of the EMI filter for DM noise. Interleaving can provide input
Fig. 4-23 shows both interleaved three-level single-stage converters used to provide input current
ripple cancellation. The power stage is duplicated and the input filter is designed at the system
at the side-band frequencies of 2fs, more specifically at 2fs±fr, where fs is the switching frequency
S1
Tr Lo
Lr A
S1
Co
Lr
Lb
S2 S2
B
A
S3
S3 Cb
B
S4 S4
S1 Lr S1
Tr Lo
A
S2
Lr
Lb
B
S2
A
S3
Cf S3
Cf
Cb
S4
B
S4
(a)
(b)
Fig. 4-23. Interleaved single-stage converters: (a) TL-PS and (b) TL-AS.
The amplitude of the relevant high-frequency harmonics to be attenuated is shown in Fig. 4-24
for both single-stage non-interleaved and interleaved systems. The total system power for both
cases is 6kW, which results in 3kW per converter for the interleaved system and 6kW for the
single converter used in the non-interleaved approach. As can be seen in Fig. 4-24, the odd
harmonics are cancelled out in the interleaved systems, while the even harmonics present the
same amplitude as those of the non-interleaved system. Another observation is that the
interleaved single-stage TL-PS converter generates lower high-frequency harmonics than its
counterpart single-stage TL-AS topology. Although the high-frequency harmonics have been
provided for the non-interleaved converters, the following comparison will consider only the
interleaved case.
0.5
0
0
0 0.1 0.2 0.3 0.4 0.5 0.6
0 0.1 0.2 0.3 0.4 0.5 0.6
Duty cycle
Duty cycle
(b)
(a)
7 3
Harmonics of switching freq. (A)
1 0.5
0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
Duty cyle Duty Cycle
(c) (d)
Fig. 4-24. Harmonics of switching frequency: (a) non-interleaved 6kW TL-PS, (b) interleaved 6kW
TL-PS, (c) non-interleaved 6kW TL-AS, and (d) interleaved 6kW TL-AS.
Fig. 4-25 and Fig. 4-26 illustrate the calculation results for the size of boost and DM filter
inductors using the CISPR 22 Class B as the conducted EMI standard. The calculation results
obtained for the VDE standard will not be shown because the two previous chapters have
demonstrated that interleaving is rather beneficial in reducing the DM filter size when the VDE
standard is taken into account. The cases compared in the following two figures include: (1) the
VIENNA rectifier (benchmark), (2) the two-switch interleaved rectifier, (3) the single-stage TL-
200
100
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
100 200
Two-switch interleaved Two-switch interleaved
TL-PS interleaved TL-PS interleaved
Filter inductance L1 [µH]
50
20
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(c) (d)
200 800
Core + Cu filter weight per phase [g]
100 400
50 200
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(e) (f)
Fig. 4-25. Design results: (a) boost inductance, combined boost inductor core + Cu weight per
phase, (c) filter inductance L1, (d) filter inductance L3, (e) filter core weight needed to implement
Ld+L1+L3, and (d) Core + Cu weight per phase required to implement Ld+L1+L3.
estimation of the total weight of the boost inductors per phase. For the interleaved cases, Fig.
4-25(a) shows the value of each inductance, whereas Fig. 4-25(b) represents the weight of the
The amount of filter inductance needed per phase is shown in Fig. 4-25(c) and Fig. 4-25(d). All
the interleaved systems require about the same amount of filter inductance per phase. Although
the interleaved systems show advantages above 150kHz, the frequency range that is most
suitable for the single-stage converters in from 40kHz to 70kHz, since the efficiency of the
single-stage topologies is very low at high switching frequencies. For 40kHz to 70kHz, the
VIENNA rectifier results in the least amount of filter inductance required to attenuate the DM
noise. Fig. 4-25(e) and Fig. 4-25(f) confirm the fact that the filter required by the VIENNA
Although the VIENNA rectifier results in the smallest DM filter size, adding the size of the boost
inductor to the size of the input filter in each phase changes the comparison. As shown in Fig.
4-26(a), the magnetic size for the combined boost and filter inductors in the switching frequency
from 40kHz to 70kHz is decreased for the interleaved cases, while Fig. 4-26(b) includes the Cu
weight in the comparison as well. According to Fig. 4-26(b), the region that most greatly reduces
the boost and filter inductor size requirements is from 50kHz to 70kHz of switching frequency.
250
TL-AS interleaved TL-AS interleaved
VIENNA 600 VIENNA
200
500
150
400
100
300
50 200
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
Fig. 4-26. Combined filter and boost inductor core weight per phase and (b) total core + Cu weight
needed to implement the boost and filter inductors per phase.
The conditions at which interleaving the single-stage converters helps to cancel the DC output
current ripple must still be determined. Cancellation of the output current ripple is beneficial to
reducing the amount of output filter capacitance required to limit the output voltage ripple to
Fig. 4-27 and Fig. 4-28 illustrate how the output current ripple is canceled out when single-stage
converters are interleaved. Fig. 4-27(a) shows both waveforms applied across points A and B
when two TL-PS converters are interleaved (the points A and B are shown in Fig. 4-23(a)). As
can be seen, the voltages across points A and B are symmetrical and phase shifted by 180o.
Because of the symmetry of the waveforms, the rectified voltages applied across the input of the
DC output filter are in phase with respect to each other. Therefore, the output filter inductor
current ripples will not be interleaved. In fact, the output current ripple cancellation in
interleaved single-stage TL-PS converters can only be achieved if more than two converters are
interleaved single-stage TL-AS converters. The voltages applied across points A and B are
shown in Fig. 4-28(a) (the points A and B are shown in Fig. 4-23(b)). It can be seen from Fig.
4-28(a) that the primary voltages applied across points A and B are phase-shifted by 180o. These
voltages are rectified and the waveforms applied across the output inductors are shown in Fig.
4-28(b). Differently from the interleaved TL-PS converters, the rectified waveforms in the TL-
AS topologies are also 180o apart, which is a necessary condition to guarantee a phase-shift
between the rectified voltage waveforms, as shown in Fig. 4-28(b). Therefore, the output
inductor current ripples will be phase-shifted and interleaved to provide output current ripple
In summary, the TL-PS AC/DC converter will cancel out the output current ripple if and only if
more than two converters are interleaved together. On the other hand, the single-stage TL-AS
topology can provide output current ripple cancellation for any number of interleaved converters.
For telecom applications, it is important to limit the output noise voltage of front-end converters
to avoid disturbance pickup in voice channels. Power supply manufacturers usually specify
weighted C-message and psophometric noise for DPS telecom applications. There are several
factors that affect the noise performance of rectifiers, such as the voltage regulation bandwidth
Ts/2
Ts/2
(a)
Ts
(a)
Ts/2 Ts/2
(b)
(b)
Ts Ts
Ts/2
Ts/2
(c)
Ts
(c)
Fig. 4-27. Output current ripple cancellation in Fig. 4-28. Output current ripple cancellation in
interleaved TL-PS converters: (a) voltage vab, interleaved TL-AS converters: (a) voltage vab,
rectified secondary voltage, and (c) output rectified secondary voltage, and (c) interleaved
inductor currents. output inductor currents.
that helps to limit the noise amplitude in the frequency range of interest. In a two-stage approach,
the fast output voltage regulation of the DC/DC converter guarantees that the noise generated in
the frequency range of interest (100Hz to 5kHz) is filtered out in order to limit the C-message
noise in the output voltage. It remains to evaluate if single-stage front-end converters can
Besides the influence of the front-end converter on the noise level, battery and installation
impedances also affect the noise performance. The noise provided by power supply
manufactures is measured under controlled laboratory conditions. However, if the same front-
end converter is placed in a different installation, the noise level produced will vary according to
The objective of the following sub-sections is to demonstrate the ability of the single-stage front-
end converters to constrain noise levels. The analysis is performed via Saber simulation, and
The magnitude of the C-message weighted noise can be determined according to the following
relationship:
∑ (E n wn )
2
dBnrC = 20 log ,
n
(4-29)
24.510 −6
message) at the component frequency fn of interest, and 24.5µV is the reference noise voltage
that results in 1pW of power in 600Ω reference impedance [84] [85]. Therefore, the contributing
factors to the noise are the voltage level of the frequency components and the weighting factor at
between the noise frequency and its disturbance effect on human hearing [84], as illustrated in
Fig. 4-29.
C-message weighing factor (dB)
-20
-40
-60
10 102 103 104
Frequency (Hz)
As can be expected, the voltage control bandwidth of single-stage front-end converters plays an
important role in limiting the C-message output voltage noise. Increasing the voltage control
bandwidth certainly helps limit the output noise. In a two-stage approach, the voltage loop
bandwidth of the DC/DC converter can be increased as much as possible to tightly regulate the
output voltage, eliminating the problem of the C-message noise in the output voltage. However,
it is unclear at this point if single-stage front-end converters can be implemented with a wide
voltage control bandwidth without deteriorating the THD of the input currents. In fact, to answer
The single-stage TL-PS converter is taken into account in the following analysis. The small-
signal modeling is an important step in order to design the voltage loop control bandwidth.
Because the intermediate bus capacitor is large enough to store energy, it is assumed that both
AC- and DC-side can be dynamically decoupled for small-signal perturbations. As a result, an
equivalent three-level ZVS DC/DC converter can be used to represent the DC-side of the single-
stage TL-PS converter. Moreover, this equivalent three-level DC/DC converter can be used to
design the voltage loop of the single-stage topology. Three-level and full-bridge converters are
identical from the small-signal standpoint, and for this reason the models developed for the
phase-shifted full-bridge converters [86] can be readily applied to three-level phase-shifted ZVS
^ ^ ^
ntVind ntVin(kiiL+kvvin)
RL ^
L iL
- + -+
C
ESR
ntv^in
- +
Ro v^o
ntVin d^ ntVin (k ^i +k ^v ) ESL
iL v in
Ro Ro
1:Deff
Fig. 4-30. Small-signal model of the equivalent three-level ZVS phase-shift DC/DC converter
representing the DC section of the single-stage TL-PS converter.
To simplify the approach, the results analyzed hereafter have been obtained at a specific
operating point for the single-stage TL-PS converter, which is 220V of input phase voltage, 3kW
of output power, and 50kHz of switching frequency. At this operating point, the simulated bus
voltage is 700V, which slightly differs from the experimental results shown in Fig. 4-12 because
Rd
ki = −
nt Vin
V 1 4 nt Llk f s
(4-30) k v = I L − o (1 − D )
2
vin ,
L 4 fs Vin
Rd = 4 nt 2 Llk f s
where Vin=351V (half of intermediate bus voltage), nt=4, IL=62.5A, Vo=48V, L=5.7µH, D=0.34,
Llk=11µH and fs=50kHz. The measured series resistance of the output filter inductor was 38mH.
In order to limit the high frequency output ripple (not the C-message noise) to lower than
480mV, the equivalent series resistance of output filter capacitor has to be lower than 15mΩ. To
accomplish this low ESR, seven capacitors of the type 380XL272M063J022 (Cornell Dubilier)
were connected in parallel. The lead inductance per capacitor is estimated in 30nH, which results
The transfer function of interest is the duty ratio to the output voltage, which can be determined
from the small-signal model shown in Fig. 4-30. The expressions obtained from that model are
summarized as follows:
1
Ro // ESR + s ESL +
^
s C ^ ^
vo = nt Vin d + nt Vin k i i
1 L
s L + R L + Ro // ESR + s ESL +
s C
(4-31) .
^ 1
ESR + s ESL +
vo sC ^
=
1
i L
Ro
ESR + s ESL + + Ro
sC
determined and the voltage loop control designed for a specified crossover frequency. As
mentioned above, the objective is to verify the effect of the voltage loop control bandwidth on
the THD of the input current, as depicted in Fig. 4-31. Increasing the control bandwidth has very
little effect on the THD of the input current. This result represents an important outcome, since a
wide control bandwidth is extremely desirable to reduce the C-message noise in the output
4.8
4.7
4.7
THD (%)
4.7
4.7
4.7
4.6
0 1000 2000 3000 4000 5000 6000
Crossover frequency (Hz)
Fig. 4-31. Simulated input current THD as a function of the crossover frequency of the single-
stage TL-PS converter.
As concluded from the previous section, the voltage loop control bandwidth does not affect the
THD of the input currents. Therefore, it remains to evaluate what role the control bandwidth
plays in constraining the C-message noise across the output voltage. To proceed with this
analysis, one can assume that the single-stage TL-PS converter is used to charge a battery string
and to feed the load according to the installation previously described [85], as illustrated in Fig.
the 8h discharge rate). Each cell is connected by intercell busbars, while cables connect the ends
of both rows. Fig. 4-32(b) shows the equivalent circuit for each section of the installation.
Single-
Load
Stage
Rectifier
44'
Battery 1/0 AWG Feed
String
+- +- +- +-
15'' 15''
2/0AWG 24 01 02 11 2/0AWG
-+ -+ -+ -+
23 22 21 12
(a)
Cables
8.65mΩ 10µH
Output inductor Ro
Vo Cables
Rectified 7x2700µF
Output capacitor
Secondary
Voltage 13.3mΩ
4.3nH VB
(b)
Fig. 4-32. (a) Installation and battery string configuration and (b) equivalent circuit.
the noise in the output voltage. The intermediate bus capacitors of the single-stage TL-PS
converter have been fixed to 100µF each for all simulations, while the DC output section has
been implemented with the equivalent circuit shown in Fig. 4-32(b). The voltage loop control
bandwidth has been adjusted according to the discussion in the previous section. The closed loop
circuit has been simulated in Saber and the C-message noise determined according to (4-29) and
the weighting factor illustrated in Fig. 4-29. The C-message noise in the output voltage is shown
in Fig. 4-33(a), while the psophometric noise is shown in Fig. 4-33(b). The psophometric noise is
dBnrC −32.5
(4-32) mV pso = 10 20 ,
where mVpso is the psophometric noise in the output voltage given in mV and dBnrC is C-
As observed in Fig. 4-33, the noise in the output voltage decreases with the increase of the
voltage loop control bandwidth. As mentioned earlier, power supply manufactures usually
specify the psophometric noise below 2mV. Therefore, the single-stage TL-PS front-end
converter can easily meet the specifications for psophometric noise in the output voltage.
4.8. Benchmarking
Since the single-stage front-end converters incorporate the DC power conversion function, the
benchmark circuit for comparison shown in Table 4-2 includes a full bridge ZVS DC/DC
converter. The data shown in Table 4-2 were obtained at 220V line-to-neutral input voltage. As
power switches and the switch RMS current. The bus capacitance for hold-up time requirement
in the single-stage front-end converters also increases because the intermediate bus voltage
fluctuates according to load and input voltage variations, as opposed to the benchmark circuit
that regulates the intermediate bus voltage to 800V. In terms of output filter capacitor, if one-
stage output LC filter is used, then the single-stage front-end converters also require more
capacitors to limit the high frequency output voltage ripple. More capacitors are required because
the output inductor of the single-stage front-end converters operates closer to DCM in order to
limit the intermediate bus voltage. As a result, the main motivation to develop single-stage front-
end converters is cost reduction, which must be traded off for performance.
45 4.0
3.5
Psophometric noise (mV)
C-message noise (dBnrC)
1.0
25
0.5
20 0.0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Crossover frequency (Hz) Crossover frequency (Hz)
(a) (b)
Fig. 4-33. (a) C-message output voltage noise and (b) psophometric noise.
Topology
VIENNA Rectifier and DC/DC Single-Stage Three-Level
Single-Stage Three-Level Phase-Shift Asymmetrical (TL-AS) Converter
converter
Phase A Phase B Phase C
DC link
(TL-PS) Converter S1
Lr Tr Lo
S1 Dr1
Lo
Io
La
A Lr
Vo
Lb Co
Cb1 Dc1 S2
Cf Co
Input Filter
Lb S2
-
Vref A B
+ Lc B
PWM
- Gdc(s) Cb2
Dc2 S3 S3
Hi(s) Phase-Shift Hdc(s) Ca Cb Cc
-
+ Vo/2
+ Dr2 Cf Cb
+ Fc(s)
- + - S4
Kv X Gv(s)
+ Vo
S4
Feature
Total power (kW) 3 3 3
Switches 7 4 4
Line freq. diodes 12 - -
Fast diodes 8 10 8
Bus voltage (V) 800 1060 1060
Switch voltage (V) 400 530 530
2.8 (PFC circuit) 14.9 (outer switches) 12.9 (outer switches)
RMS switch current (A)
5.4 (DC/DC converter) 20.3 (inner switches) 16.8 (inner switches)
Intermediate cap RMS current (A) 5.6 10.9 12.9
THD (%) - 4.7 11.2
Efficiency (%) at 50kHz - 87 89
Power under two-phase operation
2 1.6 1.4
(kW)
20ms holdup time intermediate
860µF each cap 1430µF each cap 1430µF each cap
bus cap (µF)
3x2700µF (380LX272M063J022
Output filter cap (µF) 7x2700µF (380LX272M063J022 CDE) 8x2700µF (380LX272M063J022 CDE)
CDE)
Active current control Yes No No
Sensing effort High Low Low
Control complexity High Low Low
This chapter presented two approaches for single-stage converters based on three-level
topologies: (1) the single-stage TL-PS converter and (2) the single-stage TL-AS converter. The
proposed converters provide zero-voltage turn-on and reduce the voltage stress applied across the
power switches to 50% of the intermediate bus voltage. Syntheses of the switching power stages,
along with analyses, designs, experimentation and comparisons have been provided for
The single-stage TL-PS converter was the first topology derived in this chapter to perform both
PFC and DC output voltage regulation functions. The phase-shift modulation is implemented
with commercial chips, such as those used for ZVS full-bridge converters. The THD of the input
currents at full load is always lower than 7%, despite the simplicity of the circuit. The
intermediate bus voltage becomes an issue in the design because the converter is able to regulate
only the DC output voltage, while the voltage across the intermediate bus voltage fluctuates
according to load and input voltage variations. It was found that a possible way to limit the
increase of the intermediate bus voltage is to force the output filter inductor to operate in DCM.
In this case, the maximum intermediate bus voltage decreases with the decrease of the output
inductance. However, very low output inductance deteriorates the converter efficiency because
the current stress increases in the primary switches. The switching frequency plays an important
role in the overall efficiency of the AC/DC TL-PS converter. Because the input inductors operate
in DCM, the turn-off instant of the power switches occurs when the input currents reach their
peak value within the switching period. As a result, the switching losses are high, thus requiring
than 87%.
The second single-stage converter proposed in this chapter was the TL-AS topology, which was
proposed to improve the efficiency by reducing conduction loss. Indeed, the experimental results
show an improvement in the overall efficiency as compared to the single-stage TL-PS converter.
Despite improving the overall efficiency, the single-stage TL-AS topology presented a THD that
was not as good as the THD presented by its TL-PS counterpart. The other drawback presented
by the TL-AS topology is the voltage applied across the DC blocking capacitor, which requires
the use of either polypropylene or film capacitors. The transformer also needs more turns in the
primary side in order to compensate for the DC magnetizing current that builds up during the
circuit operation. Despite these drawbacks, the TL-AS topology requires a minimal number of
This chapter also explored the interleaving of single-stage AC/DC converters. It was
demonstrated for CISPR 22 Class B standard that the size of DM filter inductors is still larger
than the size required by the VIENNA rectifier (benchmark circuit). However, the interleaving of
two single-stage converters helps reduce the overall combined sizes of filter and boost inductors,
as compared to the VIENNA rectifier. The best practical switching frequency range for the
design of interleaved single-stage AC/DC converters is between 50kHz and 70kHz. A great
reduction in the overall size of boost and filter inductors occurs above 150kHz, but at this
switching frequency range the design of single-stage converters is impractical due to the reduced
efficiency.
technique to provide cancellation for the OUTPUT current ripple. More than two interleaved
converters are needed to provide output current ripple cancellation when single-stage TL-PS
converters are used, while for the TL-AS topology any number of interleaved channels results in
single-stage topologies. However, cost has been the driving force behind the developments
described in this chapter. Despite the verified improved efficiency of the TL-AS approach, the
single-stage TL-PS converter provides much lower THD, as well as lower stress to the
transformer, which makes this circuit the best choice for single-stage front-end converters. In
terms of C-message and psophometric noise levels, the TL-PS converter is able to constrain the
noise when the voltage loop control bandwidth is properly designed. As a result, single-stage
5.1. Introduction
As demonstrated in the previous chapters, the interleaving of DCM boost rectifiers is very
effective in reducing the overall size of the boost and DM filter inductors when the CISPR 22
Class B standard is taken into account for determining the required filter attenuation. For the
VDE 0871 Class B standard, the interleaving clearly reduces the size of the DM filter because
the standard starts limiting the noise at 10kHz, as opposed to the CISPR 22 that limits the first
high frequency noise at 150kHz. As a result, the reduction of the DM input filter size due to
interleaving is not achieved for CISPR 22, but it is clear that interleaving helps reduce the
requirements for magnetic components, since the overall size of boost and DM filter inductors is
Although the interleaving of DCM boost rectifiers has demonstrated advantages, all approaches
that have been presented to this point require doubling the switching power stage. As a result, the
number of components in the interleaved front-end converter is also doubled, which may incur a
cost increase and additional layout space to accommodate the devices and interconnects for the
interleaved system. To overcome this drawback, this chapter presents a simplified interleaved
single-stage converter that eliminates the need for doubling the entire switching power stage in
order to achieve input current ripple cancellation. Comparisons are also provided to illustrate the
effectiveness of the novel interleaving technique. The analysis starts with a single-phase circuit,
The circuit topology proposed in this chapter is illustrated in Fig. 5-1. As can be seen, the circuit
is built on a three-level structure used to reduce the voltage stress applied across the power
switches. The circuit is similar to its three-phase counterpart, the TL-PS converter presented in
chapter 4, except for the addition of the auxiliary windings Na that operate as magnetic switches.
To provide proper operation for the magnetic switches [87], the number of turns of the auxiliary
windings must be greater than 2Np. The conduction paths for the inductances L1 and L2 are
required to be the same, otherwise the input current ripple cancellation will not be effective.
Na D1 S1
C1
D3 i2
L2 S2
Cf Tr Lo
+ Lr
A B
L1 S3 Ns
Co
i1 Np
D4 Ns
S4
Na D2 C2
The following assumptions are taken into consideration for describing the operation of the
circuit: (1) the input voltage is considered constant over a switching period, (2) the diodes and
switches are ideal, and (3) the inductance Lr is ignored in the analysis (the impact of the
waveforms are shown in Fig. 5-2. The switches are operated with 50% duty cycle, and phase-
shift modulation is used to control the DC output voltage. No dead time can be seen between the
gate signals shown in Fig. 5-2. Evidently, in practical applications, a small dead time is required
S1
S4
S2
S3
vAB Vbus/2
iL1
iL2
iL1 + iL2
to t1 t 3 t4
Stage 1 (to, t1) - Fig. 5-3(a): At instant to, the switch S2 is turned on and then power is transferred
to the output. As a result, the current in inductor L1 will increase linearly during this stage. Due
to the polarity of the auxiliary windings, the current through inductor L2 will be reset by the
peak input phase voltage. As a result, the current in inductor L2 is fully reset during this stage.
Stage 2 (t1, t2) - Fig. 5-3(b): At instant t2, not shown in Fig. 5-2, the switch S1 is turned off, and
the energy stored in inductor L1 is used to charge and discharge the intrinsic capacitances of S1
and S4, respectively. This voltage transition is not shown in either Fig. 5-2 or Fig. 5-3, but when
the discharge is complete, the body diode of S4 will conduct to finalize the zero-voltage
transition for S1 and S4. The conduction of the body diode of S4 slightly discharges the clamping
capacitor Cf until the voltage applied across S1 reaches the voltage across the bus capacitor C1,
thus turning on the diode Dc1. Since the body diode of S4 conducts during this stage, the switch
S4 can be turned on under ZVS. During this stage both currents through L1 and Lo are partially
reset. While the voltage difference (Vbus/2-|vin|) resets L1, the output voltage is used to reset Lo.
The voltage across the auxiliary windings is zero because the voltage applied across points A and
B is also zero.
Stage 3 (t2, t3) - Fig. 5-3(c): During this stage, the clamping diode Dc1 conducts, while the
currents through L1 and Lo continue to be reset. The voltages across the auxiliary windings
continue to be zero because vAB is shorted out by the clamping diode Dc1. This stage is finished
Stage 4 (t3, t4) – Fig. 5-3(d): When the switch S2 is turned off, the boost inductor current flowing
through L1 will charge and discharge the intrinsic capacitances of S2 and S3, respectively. When
the voltage applied across S3 reaches zero, the body diode of S3 will conduct to finalize the ZVS
turn on for S3. During this stage, the voltage polarity applied across points A and B is reversed,
which enables the voltage applied across the auxiliary windings to be reversed as well. As a
inductor L1 is fully reset by the voltage difference (Vbus-|vin|). The fourth operating stage ends
when the switch S4 turns off. Then, the energy stored in L2 will charge and discharge the intrinsic
This operating stage can also be used to derive the turns ratio between the primary and auxiliary
windings of the transformer. From the circuit shown in Fig. 5-3(d), the following expression can
be obtained:
To guarantee that L2 is magnetized at the same rate as L1, the constant k shown in (5-1) must be
k=1.
From the circuit diagram, the voltage applied across the auxiliary winding is given by:
N a Vbus
(5-2) k Vbus = .
Np 2
Thus, in order to provide the same conduction path for both boost inductors L1 and L2, the turns-
ratio between the auxiliary and primary windings for k=1 must be:
Na
(5-3) =2.
Np
The proposed technique for simplifying the interleaving of the single-stage TL-PS converter is
easily extended to three-phase applications. As shown in Fig. 5-4(a), the approach involves
constraints described above for the single-phase converter, such as the transformer turns ratio
between the primary and auxiliary windings, still apply for the three-phase circuit.
kVbus
+-
D 1 S1 Na D1 S1
C1 C1
D3 i2 D3 i2
S2 S2
Dc1 L2 Dc1
L2 Cf Cf Tr Lo
Tr Lo + Lr
+ Lr A B
A B
L1 L1 Ns
Ns S3 Co
S3 Co i1
i1 Np Np
D4 Dc2 D4 S4 Dc2 Ns
S4 Ns
C2 Na D2 C2
D2
+-
kVbus
(b)
(a)
kVbus
Na D1 S1
+-
C1 D 1 S1
C1
D3 i2 D3
S2 i2
Dc1 S2
L2 Cf L2 Dc1
+ Lr Tr Lo Cf Tr Lo
A B + Lr
A B
L1 Ns L1
S3 Co Ns
i1 S3 Co
Np i1 Np
D4 S4 Dc2 Ns D4 S4 Dc2 Ns
Na D2 C2 C2
+- D2
kVbus
(c)
(d)
Fig. 5-3. Operating stages: (a) first stage (to, t1), (b) first stage (t1, t2), (c) first stage (t2, t3), and (d)
first stage (t3, t4).
In the three-phase simplified interleaved converter shown in Fig. 5-4(a), the neutral of the power
system is connected to the switching power stage, which allows zero-sequence-order harmonics
to circulate in the input lines. To eliminate this problem, the circuit shown in Fig. 5-4(b) employs
three AC capacitors to generate an artificial neutral connection point in order to trap zero-
L1
L4
S2
Lo
L2
L5
S3 Lr
Co
L3 N
L6
S4
2N
(a)
2N S1
L1
L2
S2
L3 Lo
Lr
L4
S3 Co
L5
N
L6
S4
2N Tr
(b)
Fig. 5-4. Simplified three-phase interleaved converter: (a) simplified interleaved single-stage TL-PS
converter, and (b) including AC capacitors to eliminate the neutral point connection of the power
system.
5.3.1. Impact of the Resonant Inductance on the Effectiveness of Input Current Ripple
Cancellation
The resonant inductance Lr results from the series combination of the transformer leakage
inductance and any external inductance added to extend the ZVS operation of the proposed
converter. The effect of the resonant inductance is to produce a voltage drop across the primary
To evaluate the effect of the resonant inductance on the input current ripple cancellation, suppose
that the proposed simplified interleaved converter has been designed to deliver 3kW of output
power at a switching frequency of 50kHz. In this case, the boost inductances are 220µH, the
output filter inductance is 5.7µH, the transformer turns ratio from the primary to the secondary is
4:1 (main turns ratio), and the transformer turns ratio from the auxiliary windings to the primary
Fig. 5-5 and Fig. 5-6 show the simulation results obtained from the proposed circuit illustrated in
Fig. 5-4(b) using the design parameters described above. In the first simulation, the resonant
inductance was set to 2µH, while in the second case it was set to 5µH. The first observation that
can be made is that the current in inductor L4 is always lower than the current in inductor L1
because the voltage drop across the resonant inductor reduces the voltage reflected across the
auxiliary windings. The impact of the voltage drop across the resonant inductance on the
waveform of the interleaved current can be seen in Fig. 5-5(c) and Fig. 5-6(c), which show that
the input current ripple cancellation effectiveness decreases as the resonant inductance increases.
As a result, the noise at the switching frequency is not fully cancelled out because of the voltage
drop across the resonant inductance. As illustrated in Fig. 5-6(d), the noise at the switching
frequency becomes comparable to the noise at 2xfs, which was supposed to be the dominant
5 5
0 0
-5 -5
iL4 iL4
-10 -10
0 5 10 15 20 0 5 10 15 20
Time (ms) Time (ms)
(a) (a)
10
10
Interleaved current (A)
0 0
-5 -5
-10 -10
0 5 10 15 20
0 5 10 15 20
Time (ms)
Time (ms) (b)
(b)
0.6 0.8
0.67A 0.7A
0.55A
0.6
Amplitude (A)
Amplitude (A)
0.4
0.31A
0.4
0.2
0.2
0 0
0 100 200 300 400 500 0 100 200 300 400 500
Frequency (kHz) Frequency (kHz)
(c) (c)
Fig. 5-5. Simulation results for Lr=2µH: (a) Fig. 5-6. Simulation results for Lr=5µH: (a)
inductor currents, (b) interleaved current, and inductor currents, (b) interleaved current, and
(c) amplitude spectrum. (c) amplitude spectrum
current in the boost inductors connected through the same path of the auxiliary windings (L4, L5
and L6). For instance, the transformer turns ratio between the auxiliary and primary windings can
be greater than 2:1 in order to reflect a higher voltage across the auxiliary windings to
compensate for the voltage drop across the resonant inductance. Another possibility is to reduce
the inductance values for L4, L5 and L6 to increase the boost inductor currents and compensate
for the lower voltage reflected across the auxiliary windings. When the resonant inductance is
designed to extend the load range for ZVS operation, the solutions above are not so effective for
improving the input current ripple cancellation. Thus, the use of the circuit shown in Fig. 5-7
might be a better option. As can be seen, the proposed circuit employs an auxiliary transformer
designed with low leakage inductance to drive the auxiliary windings and to bypass the voltage
The addition of the auxiliary windings alters the primary current and increases the apparent
power requirement of the transformer. To show the impact of the auxiliary windings on the
increase of the transformer power ratings, consider that the simplified interleaved TL-PS
converter shown in Fig. 5-4(b) is designed such that the converter parameters are identical to
those used for the non-interleaved TL-PS converter introduced in chapter 4. The parameters used
in the comparison are illustrated in Table 4-1. The transformer turns ratio between the auxiliary
and the primary windings has been adjusted to 2.1:1 in order to compensate for the voltage drop
across the transformer’s 2µH leakage inductance (no external resonant inductance has been
Fig. 5-7. Simplified interleaved circuit using an auxiliary transformer to compensate for a large
resonant inductance used to increase the load range with ZVS operation.
Table 5-1. Components used in both interleaved and non-interleaved single-stage TL-PS
converters.
Non-Interleaved Single-Stage
Simplified Interleaved Single-
Three-Level Phase-Shift
Component/Parameter Stage Three-Level Phase-Shift
Converter
Converter
(see Fig. 4-2(b) on page 123)
Output Power 3kW 3kW
Switching Frequency 50kHz 50kHz
Boost Inductance 110µH 220µH (Each inductance)
2µH (Transformer leakage
Resonant Inductance 11µH
inductance)
Output Inductance 5.7µH 5.7µH
12/3/3(Main windings) + 25/25
Number of Turns 12/3/3
(auxiliary windings)
Clamping Capacitance 2µF 2µF
Fig. 5-8 shows the simulation results for both converters under discussion. The simulation was
performed at full load and 180V of line-to-neutral input voltage. As shown in Fig. 5-8(a), the
insertion of auxiliary windings significantly modifies the shape of the primary current. The two
cases present slightly differences in the voltage applied across points A and B. In either case, the
For the non-interleaved TL-PS converter, the primary RMS current is 16A, the secondary RMS
current is 44A, and the power delivered to the output is 3kW. For the simplified interleaved TL-
and the RMS current through the auxiliary windings is 5.8A. The power processed by the
transformer used in the simplified interleaved TL-PS converter is 3.85kW (850W are processed
in the auxiliary windings), which represents an increase of 28% with respect to the non-
interleaved TL-PS converter that delivers the same power to the load.
The same kind of comparison involving the non-interleaved TL-PS converter and the simplified
interleaved topology using the auxiliary transformer shown in Fig. 5-7 reveals that the auxiliary
transformer processes 1.3kW, which represents 43% of the power processed by the main
transformer. Therefore, adding an auxiliary transformer to deal with a large resonant inductance
is less advantageous than deriving the auxiliary windings from the main transformer. Evidently,
the latter is applicable when there is no resonant inductance in the circuit to increase the load
range under which the converter operates with ZVS. In fact, the addition of resonant inductance
is not important for increasing the ZVS load range because the energy stored in the boost
inductors is sufficient to provide ZVS for a wide load range, which indicates that the most
desirable topology for simplifying the interleaving of the input currents is shown in Fig. 5-4(b).
It remains to be assessed how effective the use of auxiliary windings is in reducing the overall
The proposed simplified interleaved single-stage TL-PS converter has been tested at 3kW and
50kHz of switching frequency. The results that follow were obtained from the simplified
interleaved TL-PS converter using a separate auxiliary transformer, even though this variation is
from the main transformer. A separate auxiliary transformer was used because the
implementation was relatively easy to do, since the same TL-PS converter implemented in
Fig. 5-8. Primary simulated waveforms: (a) transformer current for both simplified interleaved and
non-interleaved single-stage TL-PS converter, and (b) voltage applied across points A and B for
both converters.
As a result, the parameters and device part count of the circuit shown in Fig. 5-7 are IXFN44N80
for the MOSFETs, RUR30120 for the line rectifiers and clamping diodes, and HFA120MD40D
for the output rectifiers. Each bus capacitance consisted of three 1000µF/250V electrolytic
capacitors connected in series, while the clamping capacitor Cf and the AC capacitors Ca, Cb and
Cc were implemented with 2µF polypropylene capacitors. The output filter inductance was
5.7µH, while the output filter capacitor was 2x4700µF/100V, and the resonant inductance has
been adjusted to 11µH. The auxiliary transformer has been implemented on an E55/21 core using
17 turns of 6x22AWG Cu strand wires in the primary side, and 34 turns 2x20AWG of single-Cu
strand wires in the secondary side of the auxiliary transformer. The windings of the auxiliary
Fig. 5-9(a) shows the boost inductor currents measured at 3kW and 220V of line-to-neutral input
voltage. As can be seen, the current through inductor L4 is slightly lower than the current through
inductor L1 because of the voltage drop across the 2µH leakage inductance of the auxiliary
transformer, which has not been compensated for by a turns ratio greater than two to one.
Despite the difference between the boost inductor currents, the resulting interleaved current
The amplitude spectrum of the interleaved current is illustrated in Fig. 5-9(b). As can be seen,
the harmonic at the switching frequency and other odd harmonics are not fully cancelled out
because the interleaving is not ideal, as mentioned above. Although the interleaving is far from
the ideal case, the resulting harmonic at the switching frequency is only 0.42A, which represents
a fantastic reduction from the 6A measured in the non-interleaved case at the same operating
Fig. 5-9(c) shows the interleaved current for three different input phase voltages. A great deal of
input current ripple cancellation occurs, but the effectiveness of ripple cancellation deteriorates
as the input voltage increases. This problem was expected because the duty cycle is reduced as
the input voltage increases, which also reduces the effectiveness of the input current ripple
cancellation.
iL4 100
10-1
Amplitude (A)
iL1
10-2
10-3
10-4
iL1+iL4
10-5
5A/div 10-6
0 1 2 3 4 5 6 7
x 105
Frequency (Hz) (b)
(a)
180V
220V
260V
5A/div
(c)
Fig. 5-9. Experimental results obtained at 3kW and 50kHz: (a) boost inductor and interleaved
currents at Vin=220V, (b) amplitude spectrum of the interleaved current at Vin=220V, and (c)
interleaved current measured at three different input voltages.
This section evaluates the effectiveness of using the simplified interleaved TL-PS rectifier to
reduce the size of the DM input filter. Although this evaluation has already been done in the
previous chapter, it is repeated here because there is a new element in the spectrum of the
interleaved current, which is the noise at the switching frequency that is generated as result of the
were obtained by simulating the simplified interleaved converter shown in Fig. 5-4(b), designed
for 3kW of output power and 170V to 265V RMS of input voltage variation. No resonant
inductance was considered in the simulation, while the leakage inductance was set to 2µH.
Despite interleaving the input currents, one can still observe that the first and all odd harmonics
are present in the current spectrum. Therefore, the effect of the switching frequency noise on the
1.4
Hamonics of switching freq. (A)
fs
1.2 2fs
3fs
1
4fs
0.8 5fs
6fs
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4 0.5
Duty cycle
Fig. 5-11 shows the boost inductances per phase as a function of the switching frequency. For the
case of the TL-PS simplified interleaved converter, the boost inductance represented in Fig. 5-11
refers to each of the two inductances connected per phase. As can be noticed, this chapter
presents a comparison at 3kW of total power, while the previous chapters compared the
converters for 6kW of output power. As a result, the total filter capacitance is reduced 1.9µF per
phase.
Fig. 5-12 shows the comparison results as a function of the switching frequency. Although the
the analysis within this switching frequency range is because the efficiency of single-stage
converters at higher switching frequencies is very low. The network of the filter used for the
300 1400
1200
TL-PS simpl. interl. [µH]
250
Boost inductance
Boost inductance
1000
VIENNA [µH]
200
800
150
600
100 400
50 200
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
Fig. 5-11. Boost inductance vs. frequency: (a) 3kW TL-PS simplified interleaved converter and (b)
3kW VIENNA rectifier.
Fig. 5-12(a) and Fig. 5-12(b) show the filter inductances L1 and L3 as a function of switching
frequency for the CISPR 22 Class B standard. Clearly, the VIENNA is far better than the
interleaved system in terms of filtering requirements below 150kHz. Although the VIENNA
rectifier requires less filtering than the simplified interleaved converter, the curves show that
below 150kHz, the switching frequency has no effect on reducing the minimum inductances of
the filters. The estimated weight of the filter inductor per phase is shown in Fig. 5-12(c) and Fig.
5-12(d). As can be seen, it follows the same trends of the inductance values. From these figures,
the benefit of interleaving converters in reducing the size of the DM input filter is not clear.
However, as shown in Fig. 5-12(e) and Fig. 5-12(f), the combined weight of the equivalent boost
and filter inductors per phase is reduced to below 75kHz for the simplified interleaved TL-PS
single-stage converter.
200
300
150
200
100
50 100
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(a) (b)
500
400
100
300
200
50
100
0 0
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(c) (d)
Total core+ Cu weight per phase [g]
180 600
TL-PS simplif. interl. TL-PS simplif. interl.
VIENNA
Core weight per phase [g]
140
400
120
300
100
80 200
60 100
0 50 100 150 200 0 50 100 150 200
Switching frequency [kHz] Switching frequency [kHz]
(e) (f)
Fig. 5-12. Results of comparison: (a) filter inductance L1, (b) filter inductance L3, (c) filter core
weight, (d) combined filter core and winding weight per phase, (e) core weight of filter and boost
inductors per phase, and (f) overall filter and boost inductors weight per phase (core + winding
Cu).
A simplified interleaving technique was presented in this chapter. Based on the use of magnetic
switches, this technique is implemented without doubling the switching power stage. The
magnetic switches used in the topology are realized with auxiliary windings linked to the main
transformer. As demonstrated, the number of turns in each auxiliary winding must be at least
twice the primary number of turns to provide input current ripple cancellation.
Each input phase is connected to two boost inductors. The switching power stage and the
magnetic switches produce two high-frequency voltages that are 180o apart in a switching period.
As a result of these two phase-shifted high-frequency voltages, the two boost inductor currents
connected to the same phase are also phase-shifted by 180o, providing input current ripple
cancellation.
The influence of the resonant inductance on the input current ripple cancellation has also been
presented. The resonant inductance is used to increase the load range under which the converter
operates with ZVS. However, the effect of the resonant inductance is to generate a voltage drop
across the voltage imposed by the magnetic switches on the boost inductors. As a result, the
input current ripple cancellation effectiveness is reduced, and there is difference between the
The effect of the resonant inductance in increasing the load range for ZVS is not so important
because the input boost inductor currents are able to contribute with much more energy to the
ZVS operation than the resonant inductance itself. As a result, the resonant inductance can be
eliminated such that the only inductance connected in series with the transformer is the leakage
inductance, reducing the effect of the voltage drop across the auxiliary windings, and
auxiliary windings increases the power processed by the transformer by 28% as compared with
Regarding the input filter size, a comparison has been provided between the simplified
interleaved converter and the VIENNA rectifier, which was redesigned for an output power of
3kW. For the CISPR 22 Class B standard, the VIENNA rectifier results in smaller filter size as
compared to the simplified interleaved converter. However, when the weight of the filter
inductors is combined with the weight of the boost inductor, the simplified interleaved converter
results in a smaller overall size as compared to the VIENNA rectifier. The comparison is
meaningful for the switching frequency range between 40kHz and 75kHz, since above this
Distributed power systems are widely used to supply energy to modern telecommunication
systems, mainframe computers and servers. The major benefits of DPSs are the ability to supply
large systems, while increasing reliability, system redundancy and availability. The flexibility to
expand the system capacity as the load requirements increase and the ability to quickly replace
faulty modules are additional advantages of the DPSs. The front-end converters, built to provide
PFC function and to regulate the DC distribution bus voltage, are the DPS building blocks.
The PFC function is required in the European Community because of the harmonic standards,
whereas in the U.S. the market itself has imposed this type of feature to the front-end converters
for DPS applications. Despite the requirements and the market demands, the end user is not
willing to pay for any extra cost that has to be added in order to guarantee the PFC function. As a
result, there are many reasons for developing research on low-cost PFC circuits used for high-
Based on the previous discussion, this dissertation presented simple and low-cost solutions for
three-phase PFC used in DPS applications. The main purpose of this dissertation was to develop
3kW to 6kW PFC circuits, while achieving the following features: (1) reduced complexity and
CHAPTER 1 presented an overview of the main solutions for front-end PFC converters. From
that discussion, two circuits were selected as the benchmark circuits: (1) the CCM boost and (2)
the VIENNA rectifiers. These circuits are the most widely industry used solutions to provide
boost rectifier in the PFC stage. The advantages and drawbacks were analyzed, while
DCM boost rectifier, such as the harmonic injection method used to reduce the THD of the input
currents. When the harmonic injection method was used, 8kW could be extracted from the
single-switch DCM boost rectifier, while still complying with the IEC 61000-3-2 harmonic
standard without increasing the intermediate bus voltage beyond 800V. Because the intermediate
bus voltage was 800V, IGBTs operating at 40kHz were used in the implementation, whereas a
ZCT soft-switching circuit was implemented to allow soft turn-off for the main IGBTs. The
DCM operation of the single-switch boost rectifier requires large filters to attenuate the high
input current ripple. Therefore, interleaving the single-switch DCM boost rectifiers was explored
as an alternative method to reduce the filtering requirements. Considering the VDE 0871 Class B
EMI standard, the interleaving of two single-switch DCM boost rectifiers reduced the filtering
requirements to a level similar to that of the CCM boost and VIENNA rectifiers. On the other
hand, the reduction in filtering requirement due to interleaving is not clear when the CISPR 22 is
taken into account. To demonstrate the advantages of interleaving DCM boost rectifiers under
the CIPSR 22 EMI standard, both filter and boost inductors should be combined in the overall
magnetic size comparison. Combining the sizes of boost and filter inductors demonstrated that
interleaving is quite advantageous in reducing the requirements for magnetic devices, especially
in the switching frequency range from 50kHz to 75kHz, as well as above 150kHz. Including the
A two-stage front-end converter was developed in chapter 2 to supply 6kW of output power to a
48V DC bus. Because the intermediate bus voltage provided by the interleaved PFC system was
800V, the front-end DC/DC converter was implemented with a ZVZCS three-level converter
operated at 100kHz. The resulting system combines simplicity and performance in a compact
two-stage front-end converter. The overall system efficiency, including the EMI filter, was
90.5% at full load and nominal input voltage (220V line-to-neutral input voltage).
The second system described in CHAPTER 3 was also a two-stage front-end converter in which
the PFC stage was implemented with a two-switch three-level DCM boost rectifier. The
motivation for this particular work was to use power MOSFETs in the PFC stage. The use of
MOSFETs allows the switching frequency to be increased in order to reduce the size of the input
filter. The two-switch three-level DCM boost rectifier resulted in 8.8% THD at 220V input line-
to-neutral voltage. The harmonic injection method had a major impact in reducing the THD to
below 10% over the entire input voltage variation. Using 600V power MOSFETs with low Rds-on
achieved a PFC efficiency of 97.4% at a switching frequency of 40kHz and 96.4% at 70kHz.
Interleaving was also explored as an alternative way to decrease the input current ripple, and
consequently the filtering effort. Under the CISPR 22 EMI standard, two-channel interleaved
two-switch three-level DCM boost rectifiers had a great impact on the overall size reduction of
the boost and filter inductors. From the point of view of reduced overall boost ad filter inductors
sizes, the best switching frequency range to design the interleaved two-switch DCM boost
boost rectifiers and the DC/DC three-level ZVZCS converter yielded an efficiency of 92.1%
when the PFC stage was operated at 40kHz, and 91.3% when operated at 70kHz. As a result, the
two-switch three-level DCM boost rectifier outperforms its single-switch counterpart in terms of
both efficiency and THD. Thus, the two-switch three-level DCM boost rectifier makes the best
cost tradeoff PFC option for DPS front-end converters. The two-stage front-end converter
approach is quite suitable for telecom applications because the DC/DC converter can provide fast
regulation and easier compliance with output ripple limits and noise, such as the psophometric
noise.
Power modules for mainframe computers and server applications are usually designed for less
than 1.2kW, but with the increasing demand for computer applications, the power requirements
are likely to increase in the future. To further simplify and lower the cost of three-phase front-
end converters, the second part of this dissertation was dedicated to single-stage three-phase
front-end converters operated at zero-voltage switching. The first single-stage module presented
in CHAPTER 4 was based on the integration of a two-switch three-level DCM boost rectifier
with a three-level phase-shift ZVS DC/DC converter. The resulting single-stage three-level
phase-shift converter presented a THD lower than 7% at full load and 87% of overall efficiency,
including the input filter. To reduce the conduction loss, another single-stage front-end converter
was proposed, based on the functional integration of a two-switch three-level DCM boost
rectifier and a three-level DC/DC asymmetrical converter. This reduced the conduction loss by
50% when compared to the single-stage three-level phase-shift module. By reducing the
50kHz.
Despite improving the overall efficiency, the single-stage three-level asymmetrical topology
presented higher THD as compared to the single-stage three-level phase-shift converter. The
other drawback presented by the asymmetrical topology was the voltage across the DC blocking
capacitor, requiring the use of either polypropylene or film capacitors. The transformer also
needs more turns to compensate for the DC magnetizing current that builds up during the
asymmetrical operation of the circuit. As a result, the three-level phase-shift converter is still the
and servers. For telecom applications, it was verified that single-stage front-end converters also
provide low levels of psophometric noise in the output voltage. Both single-stage topologies
discussed in chapter 4 experience a voltage stress across the intermediate bus voltage. However,
the three-level topologies used to implement the single-stage converters help alleviate the voltage
applied across the power switches. In general, the performance of a two-stage front-end
converter is superior when compared to single-stage topologies. However, cost has driven the
development of single-stage converters. In many cases, customers are willing to trade off
Chapter 4 also explored the interleaving of single-stage AC/DC converters. For the CISPR 22
Class B EMI standard, it was demonstrated that the combined sizes of boost and filter inductors
are reduced as compared to the requirements of the VIENNA rectifier. The best practical
switching frequency range for the design of interleaved single-stage AC/DC converters is
requirements. However, interleaving single-stage rectifiers must rely on doubling the switching
approach using the single-switch three-level phase-shift converter. The technique implemented
in chapter 5, combined the power stage and magnetic switches to produce two high-frequency
the two boost inductor currents connected to the same phase were also phase-shifted by 180o,
thus providing input current ripple cancellation. The effect of the resonant inductance on the
input current ripple cancellation was investigated, and measurements to correct the problem were
presented. Among the correction measures are the elimination of the resonant inductance from
the circuit or the use of a separate auxiliary transformer to implement the auxiliary windings. As
demonstrated, the power processed by the transformer with auxiliary windings increases by 28%,
while using a separated transformer requires it to be designed to withstand 43% of the output
power. Regarding the input filter size, a comparison for the CISPR 22 EMI standard
demonstrated that the VIENNA rectifier resulted in smaller filter size, as compared to the
simplified interleaved converter. However, when the weight of the filter inductors is combined
with the weight of the boost inductors, the simplified interleaved converter results in smaller
overall magnetic size. The comparison is meaningful for the switching frequency range from
40kHz to 75kHz, since above this switching frequency the efficiency of the single-stage
converter is questionable. Because of the simplicity of canceling the input current ripple and the
low-cost candidate for mainframe computers, server, and even telecom applications.
Introduction
Kool Mµ powder cores are naturally suited to the design of designing inductors because they
provide a higher energy storage capability than gapped ferrite cores of the same size and
effective permeability. The high flux density and low core losses make Kool Mµ powder cores
quite suitable for PFC applications. Where a significant current ripple may contribute to heat
rise, such as in DCM rectifiers, Kool Mµ powder cores will present superior performance as
compared to iron powder cores as well. The following paragraphs are dedicated to describing the
procedure used to design the boost and filter inductors for the comparisons of the various
The first step for designing inductors is to define the core geometry, as illustrated in Fig. I- 1. In
the following formulae, the variable OD stands for outside diameter, ID for inner diameter, and
HT for core height. The inductor design yields continuous results, which means that OD, ID and
HT can assume any value within a certain range, which is limited by the following form-factor
constants:
OD
kd =
ID
(I-1) ,
HT
kh =
OD − ID
instead they have been based on the average form-factor constants obtained from commercial
cores [58].
OD
ID
HT
Several constraints were assumed for the design of the inductors, such as maximum current
density Jmax=450A/cm2 and maximum core temperature rise ∆T=50oC. The 60µ core material has
been chosen for the design of the inductors because it provides high saturation flux, while still
maintaining a reasonable permeability. Fig. I- 2 shows the flux density and initial permeability
The design objective is to determine the core geometry in order to calculate core and winding
weights for the purpose of comparison between the various circuits discussed in the dissertation.
A simple computer algorithm has been developed to design the inductors, which takes into
account core and winding losses. The skin effect in the conductors has been considered in the
calculations, while any proximity effect has been ignored. The following paragraphs describe the
1
Flux density [Tesla]
Initial permeability
1 0.8
0.6
0.5 0.4
0.2
0 0
2 3 4 5 2 3 4 5
10 10 10 10 10 10 10 10
Magnetic field [A/m] H [A/m]
(a) (b)
Fig. I- 2. (a) Flux density and (b) initial permeability as a function of the magnetic field for 60µ core
material.
The first design step is to determine the core area product according to the well-known
L I rms I pk
(I-2) Ap = ,
k w B pk J max
where Irms is the RMS current through the inductor, Ipk is the inductor peak current, kw=0.4 is the
window factor utilization, Bpk is the peak flux density at the peak current, and Jmax is the
For the initial design guess, the magnetic field at the peak current is set to 135 Oe5, while the
5
Multiply Oe by k1=79.577 to obtain A/m.
where µini is the initial permeability obtained from Fig. I- 2(b) for a given H, and µo=4πx10-7
The core geometry can then be determined from the area product as follows:
HT =
2
π
4
[A p ( ) ]
k h 3 k d 2 − 2k d + 1 π 3
1
4
k d HT
(I-4) OD = .
k h (k d − 1)
HT
ID =
k h (k d − 1)
The core volume and magnetic length can be determined with the help of the following
relationships:
OD 2 ID
2
(I-5) Volcore = HT π − π , and
2 2
OD + ID
(I-6) lm = π .
2
H lm
(I-7) N = floor ,
I pk
Because of the approximation above, one must correct the final magnetic field and flux density
N I pk
Hc =
lm
(I-8) .
Bc = 60 µ ini µ o H c
14243
µc
To determine whether or not the winding can be fitted inside the core window, one must
determine the cross-section area of Cu wire needed to wind the core. The Cu wire needed for
I rms
(I-9) S Cu − wire = ,
J max
where SCu-wire represents the area of a single-strand Cu wire required to withstand the RMS
current of the inductor. For the DCM boost inductor, the skin effect must be take into account by
7 .5
(I-10) δ (cm) = .
fs
When the skin depth is greater than the radius of the Cu wire, it is necessary to parallel several
single-strand Cu wires to obtain the required RMS current capability, while limiting the radius of
each single-strand Cu wire so that it is lower than the skin depth. Once the Cu wire has been
chosen, one can determine the core window area occupied by the winding, which is given by
where N is the number of turns, SCu-wire is the cross-section area of a single-strand Cu wire, and
nwire is the number of wires in parallel to satisfy the skin depth requirements. If the ratio between
the total Cu area and the window area is greater than kw (window utilization factor), then the core
area product must be increased, and the design must restart at step 2. The incremental increase
The core loss density is determined with the help of the Steinmetz equation, which is empirically
written as
The expression above was empirically derived under the consideration of a sinusoidal excitation.
As a result, this relationship can be directly used to design the filter inductors. For the CCM and
DCM boost inductors, however, the current ripple plays an important role in the core loss, and
Fig. I- 3 shows the boost inductor current for a given operating point. It is assumed that the core
is operated in the non-saturation region of the BxH curve. A Fourier expansion of the arbitrary
ripple current can be obtained for each switching period, and thus the core loss density for that
switching period can be determined by calculating the loss density for each harmonic, and by
superimposing the various harmonics yields on the total loss density for that switching period
[89] [90]. The total average core loss density over the line period is obtained by averaging the
component can also be superimposed onto the high-frequency ripple loss density. However, this
loss component is much lower than the loss caused by the ripple current. This method used to
determine the core loss density holds for linear systems only, and is invalid for non-liner
magnetic materials [91] [92]. After determining the core loss density using this procedure, the
Pcore = Pw Volcore ,
where Pw is the loss density and Volcore is the core volume determined from (I-5). The same
Fundamental
Imax
Ripple
Imin
The Cu loss is related to the winding resistance, which is in turn related to the total length of Cu
wire used in the winding. A simple representation of the winding built around the core is
depicted in Fig. I- 4(a). In order to determine the total Cu wire length, one must first calculate
rL
x HT
ID
OD
(a)
(b)
Fig. I- 4. (a) Winding dimensions and (b) cross-sectional view of core and Cu winding.
Fig. I- 4(b) shows a cross-section view of the core and winding. As can be seen, the internal
Aw − N S Cu − wire n wire
rL =
π
(I-13) ,
2
ID
Aw = π
2
The average turn length can be easily obtained from the inductor cross-section view shown in
ID OD − ID ID
(I-14) lt = 2 HT + − rL + 2 + − rL .
2 2 2
The following relationship can then be used to determine the winding loss:
RΩ / m lt N
(I-15) PCu = I rms 2 ,
n wire
The temperature rise in a wound core depends on the Cu and core losses, as well as on the total
exposed surface area of the inductor [58]. It can be determined with the empirical expression
described below:
0.833
(P + Pcore )10 3
(I-16) ∆T o = Cu ,
C
S total
where Stotal is the total surface area for heat transfer given in cm2. Stotal can be determined by the
following relationship:
π
(I-17) S total = (OD + ID )(OD + 3 ID − 8 rL + 2HT ) .
2
constrained temperature rise, then the design procedure can continue. However, if ∆T exceeds
the specified temperature rise, then one must determine which loss component is higher so that
To calculate both core and Cu weights, the total volume of the assembly must be determined
The core volume can be calculated from (I-5), while the Cu volume is simply given by the
difference (Vtotal-Volcore).
∆T
Step 1 Step 1
Reduce Reduce
Hc Jmax
N
Y Pcore >
∆T < spec
PCu
N
Y
Continue
For the weight calculation, the Kool Mµ core density is Dcore=5781Kg/m3, while the Cu density is
DCu=8290Kg/m3. The following expressions can be used to determine the core and Cu weights.
Introduction
This appendix describes how the DM filter parameters were designed for the various PFC
circuits discussed throughout the dissertation. The design guidelines are based on the results
The basic configuration of a PFC circuit and filter can be seen in Fig. II- 1. Using the phasor
diagram, the PFC circuit produces a current ia in phase with the voltage va. Because the voltage
drop across L is very small at the line frequency, one can assume that va=vin. The voltage va
imposes a 90o-leading current through C, as shown in the phasor diagram. The phasor diagram
iin L ia
Iin
ic PFC IC
- vin +
- va +
C Circuit θ
Ia Va=Vin
2 π f r C V pk
(II-1) θ = tan −1 ,
I pk
where Ipk is the fundamental peak of the current ia, and Vpk is the peak input phase voltage. From
where IDF is the input displacement factor, which is measured as the cosine of the phase angle
between the voltage vin and current iin, Ipk-min is the peak value of the current ia at minimum or
partial load, and Vpk-high is the high-line peak input voltage. For the various PFC circuits
discussed in the dissertation, the displacement factor was assumed to be 0.95, measured at high-
line input voltage and 15% of full load. For the designs discussed in the dissertation, the full
power per phase was 2kW, while the high-line input phase voltage was 265V RMS. As a result,
the maximum filter capacitance from (II-2) is approximately 3.8µF per phase. For a multiple-
stage filter, the total capacitance is the parallel combination of the various capacitances required
in each stage.
Besides the requirements for displacement factor, the filter designed for PFC applications must
not interact with the closed-loop system of the PFC circuit [88]. The latter requirement is not an
issue for DCM rectifiers because the crossover frequency for these converters is well below the
It is important to obtain small filter component values and sizes. As a result, the filter corner
frequency must be close to the frequency of the noise to be attenuated, which requires the filter
to have a very steep attenuation characteristic near the noise frequency. Only high-order filters
are able to realize this type of characteristic while maintaining a reasonable size and meeting
PFC requirements. One type of filter that is known to provide such characteristic is the Cauer-
Chebyshev (CC) filter, also known as elliptic-integral filter [57]. A two-stage CC filter with
previous work [57] (pages 200-201), and are shown in Table II- 1.
L1n L3n
L2n C4n
Req1=1 Req2
C2n
LISN DM Filter
PFC
LISN DM Filter
Circuit
LISN DM Filter
(a)
LISN DM Filter
Rd
50µH 1 L1 L3
2 3
Ls Ld
Cp L2 C4
Inoise
Vnoise 50Ω Rp C2
(b)
Fig. II- 3. (a) Three-phase setup and (b) equivalent circuit per phase.
LISN is shown in Fig. II- 3(a). As can be seen, the neutral point is grounded, while three single-
phase LISNs are used to measure the noise. The calculation of the filter parameters is performed
per phase, using the equivalent circuit illustrated in Fig. II- 3(b). The parameters Rd and Ld are
the high-frequency damping resistor and the low-frequency current bypass inductor, respectively.
In the design of the input filter for all converters described throughout the dissertation, a 6dBµV
margin was used, and no common-mode noise was ever considered in the calculation. The
following paragraphs describe a computer algorithm used to design the input filter parameters.
Define the standard to be used for the design of the input filter. As can be seen throughout this
dissertation, the VDE 0871 and the CISPR 22, both Class B, have been selected to design and
Determine the worst-case scenario for the input current ripple generated by the converter under
consideration. Determine the spectrum of the worst-case input current ripple to be attenuated by
the DM filter.
Choose the normalized filter parameters from Table II- 1. The design should start with the filter
a) Reference Frequency
2 π f noise
(II-3) ω r = 0.8 ,
Ωs
where Ωs is the normalized frequency from Table II- 1 and fnoise is the frequency of the noise to
be attenuated. The reference frequency is chosen to be below the noise frequency in order to
reduce the effect of the filter parameter variations that occur due to temperature, tolerance, aging,
etc., thus guaranteeing that the noise frequency will stay within the filter stop-band region.
C 2n + C 4n
(II-4) Rd = ,
ω r C max
where Cmax is obtained from (II-2) at high line and 15% of full load.
Lin Rd
(II-5) Li = , where i = 1, 2 and 3 .
ωr
Lkn Rd
(II-6) Ck = , where k = 2 and 4 .
ωr
The inductor Ld solves the dissipation problem in the damping resistor Rd by providing an
alternative path for the line-frequency current. For the parallel network Rd//Ld to be effective, the
corner frequency fd=Rd/(2πLd) must be lower than the lowest filter frequency pole. As a result,
In this step, the design is theoretically verified by using the equivalent network shown in Fig. II-
3(b). The node equations for the system shown in Fig. II- 3(b) are given below:
0
0 =
I noise
1 1 1 1
s L + 1
+
R sL
−
R sL
0
(II-7) s Rp + s L1 + d d s L1 + d d .
s C p R d + s Ld R d + s Ld V
1 1 1 1 1 1
− + + − V2
R sL R d s Ld 1 s L3 s L3
s L1 + d d s L1 + s L2 + V
R d + s Ld R d + s Ld s C2 3
1 1
0 − + s C4
s L3 s L3
The system above can be solved for the node voltages, and consequently for the noise measured
The design is complete if the noise across the LISN resistors is 6dBµV below the limit
established by the EMI standard under consideration. Otherwise, a new set of normalized
parameters that provide more attenuation must be selected from Table II- 1, and steps 4 and 5
must be repeated.
[1] W.A. Tabisz, M. Jovanovic and F.C. Lee, “Present and Future of Distributed Power
Systems,” in APEC 1992, pp. 11-18.
[2] F. Bodi, “Distributed Power Systems (Telecommunication Power Supply),” in INTELEC
1988, pp. 143-150.
[3] C.C. Heath, “The Market for Distributed Power Systems,” in APEC 1991, pp. 225-229.
[4] X. Zhou, “Low-Voltage High-Efficiency Fast-Transient Voltage Regulator Module,” Ph.D.
Dissertation, Virginia Tech, July 1999.
[5] P-L. Wong, “Performance Improvements of Multi-Channel Interleaving Voltage Regulator
Modules with Integrated Coupling Inductors,” Ph.D. Dissertation, Virginia Tech, March
2001.
[6] IEC 61000-3-2, “Electromagnetic Compatibility (EMC) – Part 3-2: Limits for Harmonic
Current Emissions (Equipment Input Current ≤ 16A per Phase),” Edition 1.2, 1998.
[7] IEC 61000-3-4, “Electromagnetic Compatibility (EMC) – Part 3-4: Limitation of Emission
of Harmonic Current in Low-Voltage Power Supply Systems for Equipment with Rated
Current Greater Than 16A,” First Edition, 1998.
[8] IEEE Standard 519, “IEEE Recommended Practices and Requirements for Harmonic
Control in Electrical Power Systems,” 1993.
[9] D. Chapman, D. James and C. J. Tuck, “A High Density 48V 200A Rectifier with Power
Factor Correction,” in INTELEC 1993, pp. 188-125.
[10] M.L. Heldwein, A.F. Souza and I. Barbi, “A Simple Control Strategy Applied to Three-
Phase Rectifier Units for Telecommunications Applications Using Single-Phase Rectifier
Modules,” in PESC 1999, pp. 795-800.
[11] F. Alecks, M. Ford and M. Tuffs, “Design of an Advanced High Power Density 1U
Intelligent Rectifier,” in INTELEC 2000, pp. 17-23.
[12] G. Spiazzi and F.C Lee, “Implementation of Single-Phase Boost Power-Factor-Correction
Circuits in Three-Phase Applications,” in IEEE Transactions on Industrial Electronics, Vol.
44, No. 3, June 1997, pp. 365-371.
[13] J.R. Pinheiro and I. Barbi, “The Three-Level ZVS PWM Converter A New Concept in
High-Voltage DC-to-DC Conversion,” in IEEE IECON 1992, pp. 173-178.
[14] J. Pinheiro and I. Barbi, “Wide Load Range Three-Level ZVS-PWM DC-to-DC converter,”
in IEEE PESC 1993, pp. 171-177.
[15] N.H. Kutkut, G. Luckjiff and D.M. Divan, “A Dual Bridge High Current DC-to-DC
Converter with Soft Switching Capability,” in IEEE IAS Annual Meeting 1997, pp. 1398-
1405.
References 222
[16] F. Canales, P. Barbosa and F.C. Lee, “A Zero Voltage, Zero Current Three-Level DC/DC
Converter,” in APEC 2000, pp. 314-320.
[17] R. Ayyanar, N. Mohan and J. Sun, “Single-Stage Three-Phase Power-Factor-Correction
Circuit Using Three Isolated Single-Phase Sepic Converters Operating in CCM,” in PESC
2000, pp. 353-358.
[18] J.W. Kolar, H. Ertl and F.C. Zach, “Design and Experimental Investigation of a Three-
Phase High-Power Density High Efficiency Unity Power Factor PWM (VIENNA) Rectifier
Employing a Novel Integrated Power Semiconductor Module,” in APEC 1996, pp. 514-
523.
[19] J.W. Kolar and F.C. Zach, “A Novel Three-Phase Utility Interface Minimizing Line
Current Harmonics of High-Power Telecommunications Rectifier Modules,” in IEEE
Transactions on Industrial Electronics, Vol. 44, No. 4, Aug. 1997, pp. 456-467.
[20] P. Bialoskorski and W. Koczara, “Unity Power Factor Three-Phase Rectifiers,” in PESC
1993, pp. 669-674.
[21] R. Gules, A.S. Martins and I. Barbi, “A Switched Three-Phase Three-Level
Telecommunications Rectifier,” in INTELEC 1999.
[22] S. Chattopadhyay and V. Ramanarayanan, “Digital Implementation of a Line Current
Shaping Algorithm For Three-Phase High Power Factor Boost Rectifier Without Input
Voltage Sensing,” in APEC 2001, pp. 592-598.
[23] S. Fukuda and K. Koizumi, “Optimal Control of a Three Phase Boost Rectifier for Unity
Power Factor and Reduced Harmonics,” in Power Electronics and Drive Systems 1995, pp.
34-39.
[24] H. Mao, D. Boroyevich, A. Ravindra and F.C. Lee, “Analysis and Design of High
Frequency Three-Phase Boost Rectifiers,” in APEC 1996, pp. 538-544.
[25] C. Cuadros, D. Boroyevich, S. Gataric, V. Vlatkovic, H. Mao and F.C. Lee, “Space Vector
Modulated, Zero-Voltage Transition Three-Phase to DC Bidirectional Converter,” in PESC
1994, pp. 16-23.
[26] C.T. Cruz and I. Barbi, “A Passive Lossless Snubber for the High Power Factor
Unidirectional Three-Phase Three-Level Rectifier,” in IECON 1999, pp. 909-914.
[27] J.W. Kolar, H. Ertl and F.C. Zach, “Influence of the Modulation Method on the Conduction
and Switching Losses of a PWM Converter System,” in IEEE Transactions on Industry
Applications, 1991, No.6, pp.1065-1075.
[28] A. Busse and J. Holtz, “Multiloop Control of a Unity Power Factor Fast Switching AC to
DC Converter,” in PESC 1982, pp. 171-179.
[29] K. Wang, D. Boroyevich and F.C. Lee, “Charge Control of Three-Phase Buck PWM
Rectifiers,” in APEC 2000, pp. 824-830.
[30] J. Doval-Gandoy and C.M. Penalver, “Dynamic and Steady State Analysis of a Three-Phase
Buck Rectifier,” in IEEE Transactions on Power Electronics, Vol. 15, No. 6, Nov. 2000, pp.
953-959.
References 223
[31] V. Grigore and J. Kyyra, “High Power Factor Rectifier Based on Buck Converter Operating
in Discontinuous Capacitor Voltage Mode,” in IEEE Transactions on Power Electronics,
Vol. 15, No. 6, Nov. 2000, pp. 1241-1249.
[32] V. Vlatkovic, D. Boroyevich and F.C. Lee, “A Zero-Voltage Switched, Three-Phase
Isolated PWM Buck Rectifier,” in IEEE Transactions on Power Electronics, Vol. 10, No. 2,
March 1995, pp. 148-157.
[33] M. Baumann, U. Drofenik and J.W. Kolar, “New Wide Input Voltage Range Three-Phase
Unity Power Factor Rectifier Formed by Integration of a Three-Switch Buck-Derived
Front-End and a DC/DC Boost Converter Output Stage,” in INTELEC 2000, pp. 461-470.
[34] J. Salmon, “Circuit Topologies for PWM Boost Rectifiers Operated From 1-Phase and 3-
Phase AC Supplies and Either Using Single or Split DC Rail Voltage Outputs,” in APEC
1995, pp. 473-479.
[35] J. Salmon, “3-Phase PWM Boost Rectifier Circuit Topologies Using 2-Level and 3-Level
Asymmetrical Half-Bridges,” in APEC 1995, pp. 842-848.
[36] J. Salmon, “Operating a Three-Phase Diode Rectifier with a Low-Input Distortion Using a
Series-Connected Dual Boost Converter,” in IEEE Transactions on Power Electronics, Vol.
11, No. 4, July 1996, pp. 592-603.
[37] E. Mehl and I. Barbi, “An Improved High-Power Factor and Low-Cost Three-Phase
Rectifier,” in IEEE Transactions on Industrial Applications, Vol. 33, No. 2, March/April
1997, pp. 485-492.
[38] R. Naik, M. Rastogi and N. Mohan, “Third Harmonic Modulated Power Electronics
Interface with Three-Phase Utility to Provide a Regulated DC Output and to Minimize
Line-Current Harmonics,” in IEEE Transactions on Industry Applications, Vol. 31, No. 3,
May/June 1995, pp. 598-602.
[39] P. Pejovic, “Two Three-Phase High Power Factor Rectifiers that Apply the Third Harmonic
Current Injection and Passive Resistance Emulations,” in IEEE Transactions on Power
Electronics, Vol. 15, No. 6, Nov. 2000, pp. 1228-1240.
[40] G. Joos, “Simulation of Active Power Filters Using Switching Functions,” in 6th Workshop
on Computers in Power Electronics, 1998, pp. 163-167.
[41] R.E. Beighley, C.A. Gougler and J.R. Robson, “Application of Active Harmonic Filters for
Power Quality Improvement,” in Power Engineering Society Summer Meeting 1999, pp.
1175-1180.
[42] A. R. Prasad, P. Ziogas and S. Manias, “An Active Power Factor Correction Technique for
Three-Phase Diode Rectifiers,” in PESC 1989, pp. 58-66.
[43] J.W. Kolar, H. Ertl and F.C. Zach, “A Comprehensive Design Approach for a Three-Phase
High-Frequency Single-Switch Discontinuous-Mode Boost Power Factor Corrector Based
on Analytically Derived Normalized Converter Component Ratings,” in IEEE Transactions
on Industry Applications, Vol. 31, No. 3, May/June 1995, pp. 569-582.
References 224
[44] L. Simonetti, J. Sebastian and J. Uceda, “Single-Switch Three-Phase Power Factor Under
Variable Switching Frequency and Discontinuous Input Current,” in PESC 1993, pp. 657-
662.
[45] J.W. Kolar, H. Ertl and F.C. Zach “Space Vector-Based Analytical Analysis of the Input
Current Distortion of a Three-Phase Discontinuous-Mode Boost Rectifier System,” in
PESC 1993, pp. 696-703.
[46] K. Schenk and S. Cuk, “A Simple Three-Phase Power Factor Corrector with Improved
Harmonic Distortion,” in PESC 1997, pp. 399-405.
[47] Q. Huang and F. Lee, “Harmonic Reduction in a Single-Switch, Three-Phase Boost
Rectifier with High Order Harmonic Injected PWM,” in PESC 1996, pp. 790-797.
[48] F. Lin and D.Y. Chen, Reduction of Power Supply EMI Emission by Switching Frequency
Modulation,” in IEEE Transactions on Power Electronics, Vol. 9, No. 1, Jan. 1994, pp. 132-
137.
[49] D. Perreault and J. Kassakian, “Design Evaluation of a cellular Rectifier System with
Distributed Control,” in PESC 1998, pp. 790-797.
[50] P. Barbosa and F.C. Lee, “Design Aspects of Paralleled Three-Phase DCM Boost
Rectifiers,” in PESC 1999, pp. 331-336.
[51] P. Barbosa, F. Canales, J.C. Crebier and F.C. Lee, “Interleaved Three-Phase Boost
Rectifiers Operated in the Discontinuous Conduction Mode: Analysis, Design
Considerations and Experimentation,” in IEEE Transactions on Power Electronics,
Volume: 16, Issue 5, Sept. 2001, pp. 724 –734.
[52] J.C. Crebier, M. Brunello and J.P. Ferrieux, “Differential Mode Current Harmonics
Forecast for DCM Boost Rectifiers Design,” in EPE 1999.
[53] S. Gataric, D. Boroyevich and F. C. Lee, “Soft-Switching Single-Switch Three-Phase
Rectifier with Power Factor Correction,” in APEC 1994, pp. 738-744.
[54] J.C. Crebier, P. Barbosa, F. Canales, F.C. Lee and J.P. Ferrieux, “Frequency Domain
Analysis and Evaluation of Differential Mode Input Current for Three-Phase DCM Boost
Rectifiers With Different Control Strategy,” in PESC 2000, pp. 482-487.
[55] F-Y. Shih; D.Y., Chen, Y-P. Wu and Y-T. Chen, “A Procedure for Designing EMI Filters
for AC Line Applications,” in IEEE Transactions on Power Electronics, Vol. 11, Issue 1,
Jan. 1996, pp. 170 – 181.
[56] V. Vlatkovic, D. Boroyevich and F.C. Lee, “Input Filter Design for Power Factor Circuits,”
in IEEE Transactions on Power Electronics, Vol. 11, No. 1, Jan. 1996, pp. 199-205.
[57] A.I. Zerev, “Handbook of Filter Design,” New York, Wiley and Sons, 1967.
[58] Kool Mµ Powder Cores, provided by Magnetics, Inc.
[59] J-G Cho, J.A. Sabate, G. Hua and F.C. Lee, “Zero-Voltage and Zero-Current-Switching
Full Bridge PWM Converter for High-Power Applications,” in IEEE Transactions on
Power Electronics, Volume 11, Issue 4, July 1996, pp. 622 –628.
References 225
[60] R. Redl, L. Balogh and D.W. Edwards “Optimum ZVS Full-Bridge DC/DC Converter with
PWM Phase-Shift Control: Analysis, Design Considerations, and Experimental Results,” in
APEC 1994, pp. 159 –165.
[61] B-H Cho, D-Y Lee, S-B Yoo and D-S Hyun, “A Novel Full-Bridge ZVZCS PWM DC/DC
Converter with a Secondary Clamping Circuit,” in PESC 1998, pp. 936 –941.
[62] F. Canales, P. Barbosa, J.M. Burdio and F.C. Lee, “A Zero Voltage Switching Three-Level
DC/DC Converter,” in INTELEC 2000, pp. 512 –517.
[63] F. Canales, P. Barbosa and F. C. Lee, “A Zero Voltage, Zero Current Three-Level DC/DC
Converter,” in IEEE APEC 2000, pp. 314-320.
[64] F. Canales, P. Barbosa and F.C. Lee, “A Zero Voltage and Zero Current Switching Three-
Level DC/DC Converter Using A Lossless Passive Snubber,” in CPES Seminar 2000, pp.
372-377.
[65] D.M. Xu, C. Yang, J.H. Komg and Z. Qian, “Quasi Soft-Switching Partly Decoupled
Three-Phase PFC With Approximate Power Factor,” in APEC 1998.
[66] J.C. Crebier, M. Brunello and J.P. Ferrieux, “A New Method for EMI Study in Boost
Derived PFC Rectifiers,” in PESC 1999, pp. 855-860.
[67] Ansoft PEmag, “2D Power Electronics Parameter Extraction Simulator,” 2001.
[68] “Power Conversion & Line Filter Applications,” Micrometals Iron Powder Cores, Issue I,
Feb. 1998.
[69] Power Design Tools Inc., “Simplis Technical Reference Guide,” 1994.
[70] J. Liu, W. Chen, J. Zhang, D. Xu and F.C. Lee, “Evaluation of the Power Losses in
Different CCM Mode Single-Phase Boost PFC Converters via a Simulation Tool,” in IAS
2001, pp. 2455-2459.
[71] J. Contreras and I. Barbi, “A Three-Phase High Power Factor PWM ZVS Power Supply
with a Single Power Stage,” PESC’94, pp. 356-362.
[72] K. Schenk and S. Cuk, “A Simple Three-Phase Power Factor Corrector with Improved
Harmonic Distortion,” in PESC 1997, pp. 399-405.
[73] D.C. Martins and M.M. Casaro, “Isolated Three-Phase Rectifier with High Power Factor
Using the Zeta Converter in Continuous Conduction Mode,” in IEEE Transactions on
Circuits and SystemsI: Fundamental Theory and Applications, Vol. 48, No. 1, Jan. 2001,
pp. 74-80.
[74] J. Minbock and J. Kolar, “Design and Experimental Investigation of a Single-Switch Three-
Phase Flyback-Derived Power Factor Corrector,” in INTELEC 2000, pp. 471-478.
[75] M.T., Madigan, R.W. Erickson and E.H. Ismail, “Integrated High-Quality Rectifier-
Regulators,” in IEEE Transactions on Industrial Electronics, Volume 46, Issue 4, Aug.
1999, pp. 749-758.
[76] R. Redl and L. Balogh, “Design Considerations for Single-Stage Isolated Power-Factor-
Corrected Power Supplies With Fast Regulation of the Output Voltage,” in APEC 1995, pp.
454-458.
References 226
[77] L.H. Mweene, C.A. Wright and M.F. Schlecht, “A 1kW, 500kHz Front-End Converter for a
Distributed Power System,” in APEC 1989, pp. 423-432.
[78] R. Redl, N.O. Sokal and L. Balogh, “A Novel Soft-Switching Full-Bridge DC/DC
Converter: Analysis, Design Considerations, and Experimental Results at 1.5kW, 100kHz,”
in IEEE Tran. on Power Electronics, Vol. 6, Issue 3, July 1991, pp. 408-418.
[79] A. Nabae, I. Takahashi and H. Akagi, “A New Neutral Point Clamped PWM Inverter,” in
IEEE Transac. On IA, Vol. IA-17, No. 5, Sept./Oct. 1981, pp. 518-523.
[80] T.A. Meynard and H. Foch, “Multi-Level Conversion: High Voltage Choppers and
Voltage-Source Inverters,” in PESC 1992, pp. 397-403.
[81] E. Deschamps and I. Barbi, “A Comparison Among Three-Level ZVS-PWM Isolated DC-
to-DC Converters,” in IECON 1998, pp. 1024-1029.
[82] E. Deschamps and I. Barbi, “A Three-Level ZVS PWM DC-to-DC Using the Versatile
Multilevel Commutation Cell,” in 4th Brazilian Power Electronics Conference (COBEP)
1997, pp. 85-90.
[83] P. Barbosa, F. Canales and F.C. Lee, “A Distributed Power Systems for Medium Power
Applications,” in IAS 2000, pp. 2546-2551.
[84] W. Sharper, “Powerful Noise Problems and Solutions,” in Telephony 1982, pp. 92-100.
[85] K.E. White and E.C. Rhyne, “Effect of Battery Spring Inductance on Load Circuit Noise
and Battery Charger Filter Performance,” in INTELEC 1986, pp. 499-506.
[86] V. Vlatkovic, J. Sabaté, R. Ridley, F.C. Lee and B.C. Cho, “Small-Signal Analysis of the
Phase-Shifted PWM Converter,” in IEEE Transaction on Power Electronics, Vol. 7, No. 1,
Jan. 1992, pp. 128-135.
[87] I. Takahashi and R.Y. Igarashi, “A Switch Power Supply of 99% Power Factor by Dither
Rectifier,” in INTELEC 1991, pp. 714-719.
[88] R.D. Middlebrook, “Input Filter Considerations in Design and Application of Switching
Regulators,” in IEEE IAS 1976, pp. 94-107.
[89] P.M. Gradzki, M.M. Jovanovic and F.C. Lee, “Computer Aided Design for High-Frequency
Power Converters,” in APEC 1990, pp. 336-343.
[90] R. Servens, “HF-Core Losses for Non-Sinusoidal Waveforms,” in HFPC 1991, pp. 140-
148.
[91] A. Brockmeyer, M. Albach and T. Dürbaum, “Remagnetization Losses of Ferrite Materials
Used in Power Electronic Applications,” in PCIM 1996.
[92] M. Albach, T. Dürbaum and A. Brockmeyer, “Calculating Core Losses in Transformers for
Arbitrary Magnetization Currents – A Comparison of Different Approaches,” in PESC
1996, pp. 1463-1468.
References 227
Vita
Catarina, Florianópolis, Brazil, in 1993. From October 1993 to July 1995, he was with the Power
Electronics Institute (INEP) of the Federal University of Santa Catarina, where he worked on
In August of 1995, he joined the Department of Electrical Engineering of the Federal University
of Paraná, Curitiba, Brazil. Since 1997, he has been with the Center for Power Electronics
Systems (CPES) of Virginia Polytechnic Institute and State University. He first joined CPES as a
PhD student, and in June of 2001 he became the technical director for the Center.
During his years in CPES, he was the recipient of a fellowship granted by the Brazilian Council
for Scientific and Technologic Development (CNPq) to pursue a PhD degree. His research
interests include power factor correction circuits, distributed power systems, modeling of power
converters, and integration of power electronics systems. He is a member of the Sobraep (The
Brazilian Power Electronics Society) and of the IEEE. He is a reviewer for the IEEE